Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Overview description of the Gambian GABECE Educational Data and associated algorithms and unsupervized learning process

Overview description of the Gambian GABECE Educational Data and associated algorithms and... OVERVIEW DESCRIPTION OF THE GAMBIAN GABECE EDUCATIONAL DATA AND ASSOCIATED ALGORITHMS AND UNSUPERVIZED LEARNING PROCESS y yy yyy yyyy ´ ´ OUSMANE SAINE , SOUMAILA DEMBELE , GANE SAMB LO , MOHAMED CHEIKH HAIDARA Abstract. As the first paper of a series of exploratory analysis and statistical investigation works on the Gambian GABECE data based on a variety of statistical tools, we wish to begin with a thorough unsupervised learning process through descriptive and exploratory meth- ods. This will lead to a variety of discoveries and hypotheses that will direct future research works related to this data. Ousman Saine University of The Gambia, Banjul, The Gambia & Policy Planning, Analyses, research and Budgeting Directorate, Ministry of Basic and secondary Education, Willy Thorpe Palace, Banjul, The Gambia ousmansaine@edugambia.gm yy Dr Soumaila DEMBELE Universite ´ des Sciences Sociale et de Gestion de Bamako ( USSGB) Faculte ´ des Sciences Economiques et de Gestion (FSEG) Email: soumidemlpot@gmail.com yyy Gane Samb Lo. LERSTAD, Gaston Berger University, Saint-Louis, Sen ´ eg ´ al (main affiliation). LSTA, Pierre and Marie Curie University, Paris VI, France. AUST - African University of Sciences and Technology, Abuja, Nigeria gane-samb.lo@edu.ugb.sn, gslo@aust.edu.ng, ganesamblo@ganesamblo.net Permanent address : 1178 Evanston Dr NW T3P 0J9,Calgary, Alberta, Canada. yyyy Dr Mohamed Cheikh HAIDARA Work Affiliation : Depar ´ tement de Mathematiq ´ ue , Universite ´ Cheikh Anta Diop (UCAD), Dakar, Senegal Ecole Super ´ ieure Polytechnique, Universite ´ Cheikh Anta Diop (UCAD), Dakar, Senegal Research affiliation : LERSTAD, Universite ´ Gaston Berger (UGB), Saint-Louis, Senegal Emails : chheikhh@yahoo.fr Keywords. record values; record time; endpoints of a cumulative distribution function; hit- ting times; central limit theorem, law of the iterated logarithm; Berry-Essen bound; band of consistency; confidence intervals. AMS 2010 Mathematics Subject Classification: 62G30; 60Fxx; 63H10 arXiv:2012.08488v1 [stat.AP] 15 Dec 2020 y yy yyy yyyy ´ ´ 2 OUSMANE SAINE , SOUMAILA DEMBELE , GANE SAMB LO , MOHAMED CHEIKH HAIDARA 1. Introduction In this paper, the GABECE data is first presented and next statistically studied for the first time. We proceed to an unsupervised learning pro- cess. We will provide the description of the variables and consider the geographical and temporal sub-data. For each sub-data, different tech- niques will be applied : cleaning, outliers detection, frequency and related tables, spatial (geographical) or gender patterns, proportions of grades and evolution in time, etc. A series of statistical analyzes will be undergone on the data by ourselves or other researchers. To prepare that, we wish to devote a note on the gen- eral presentation of the data and an unsupervised learning process with respect to the gender, the area and the time. Also R codes (R Core Team (2020) and Crawley (2009)) aiming at reading the data and getting the variables presented in Section 2 will be provided in Appendix II in page 21. The rest of the paper is organized as follows. We will complete this intro- duction by the main findings in this paper and related recommendations. The data is fully described in Section 2. In Section 3, the GABECE data is compared with national statistics concerning population sizes of regions and genders to see whether or not access to education is fair in the Gambia. In section 4, the completeness of the the data is studied and it it will be shown that the data present un non-negligible number of incomplete data everywhere. In Section 5, a cleaning study is undergone by the detection of the outliers and it will appear that eliminating outliers, as usually rec- ommended, would lead to a serious error since excellent students would simply ignored in the study. Instead of rejecting, we will propose regroup- ing the grades. In Section 6, we undergo a general investigation of the performances. We close the paper by conclusions in Section 7. Appen- dices are put after the bibliography. . The main conclusions are summarized below. (C1) Neither the regions nor the two genders are well represented in the educational GABECE data. The western regions, in general, have more students at grade 13 when compared with these real weighs in the coun- try. Although being more numerous in the country, females are under repre- sented in the data. Even worse, the number of females in the data behaves THE GAMBIAN GABECE EDUCATIONAL DATA 3 chaotically where the number of males steadily increases with the year. (R1) As a consequence, we strongly recommend to use sampling meth- ods with sample sizes proportional to the real weighs of sub-groups in the global population. (C2) There is a big number of incomplete data in all years, in all regions, and in each gender. After dropping them, we still have enough data to proceed to a significant analysis data. (R2) On this side, we recommend statistical studies based on missing data estimation techniques. There is a lot to do on this, using nearest neigh- borhood, regression, classification, etc. (C3) The statistical studies are about the grades in the core disciplines: Mathematics, Sciences, English, and SES. The grades range from 1 (ex- cellent) to 9 (fail). Unfortunately, the level of students is not very fairly good since the fail grade is dominant almost everywhere. As well, there is a very few number of excellent students. (R3) The result of that situation is that excellent grades 1, 2 or 3 are de- tected as outliers. So, in the statistical studies, such grades should be excluded. But, fortunately, a regrouping of all grades into three levels seems to be the best way to overcome that problem. Those three levels are: credit (regrouping credit grades 1 to 6), pass (regrouping 8 and 8) and fail. The paper is destined to be a basis for a series of statistical studies on GABECE data and beyond on similar data for countries of the West African Examinations Council, whose The Gambia is member of, and on educa- tional data in general. 2. Overall presentation of the GABECE data The West African Examinations Council was established in 1952. This Council has the responsibility of ascertaining examinations required in the public interest in the English-speaking West African countries, com- parable to those of equivalent examining authorities internationally, and to develop syllabuses, arranges and administers examinations and awards y yy yyy yyyy ´ ´ 4 OUSMANE SAINE , SOUMAILA DEMBELE , GANE SAMB LO , MOHAMED CHEIKH HAIDARA certificates. It also conducts a number international examinations. The data we are going to use in this document is from the above afore- mentioned body. The data covers the period from the 2012 academic year to 2017. The mode of the GABECE examination is that each student is required to choose 7 subjects at least and 9 subjects at most. Out of the subjects taken, four (4) are compulsory and are called core subjects, which are English (ENG), Mathematics (MATH), Sciences (SCIENCES) and Social and Environmental Studies (SES). Individual performance grants the student admission to Senior Secondary Education. The grading system in this result is categorized as follows : Credit (grades from 1 to 6), Pass (7 to 8), Fail (9). The result of a student is determined by the grades of the four core subjects plus the two best grades from the student’s choice. The best result a student could have is the aggregated grade from 6 (six), one in all the core subjects, plus one in the two choices of the students and the worst result is aggregated to 54 with failure in all subjects with grade 9. In this study, we focus on the core grades (G1), (G2), (G3) and (G4) corre- sponding to English, Mathematics, Sciences and SES respectively. Here are the conventions we will be using to know whether we use a grade for the whole population or not, for all the years or not, for both genders or not, or whether we use a grade for one particular gender, or a year or an area. The corresponding variables of the core grades are denoted as G1, G2, G3 and G4 before the cleaning of the data and cover all areas, all time and all gender. We also have the variables year (from 2012 to 2017), gender (coded 1 for males and 2 for females), and region (coded for 1 to six regions). The missing data, which correspond to absences of students or submis- sion of blank sheets by student, represent 4.54% of the data (5792 out of 121838). The new variables become : G1A, G2A, G3A, G4A, yearA, genderA, regionA. THE GAMBIAN GABECE EDUCATIONAL DATA 5 When we restrict our analysis on a specific year for all gender, the variables are: G1AY, G2AY, G3AY, G4AY, yearAY, genderAY, regionAY. When we restrict our analysis on a specific region for all genders and years, the variables are : G1AR, G2AR, G3AR, G4AR, yearAR, genderAR, regionAR. When we restrict our analysis on a specific region and for a specific year, for all genders, the variables are : G1AYR, G2AYR, G3AYR, G4AYR, yearAYR, genderAYR, regionAYR. In each of the previous cases, the specified year and region are given by the variables yearOfStudy and regionOfStudy. If we want to specify and fix a gender, we use the filtering codes : Vari- able[gender==”1”] for males (and Variable[gender==”2”] for females). For example, the variables G1AR[genderAR==”1”], G2AYR[genderAYR==”2”], G3A[genderA==”1”], respectively contain: a) the grades for males in English for the region determined by the value of regionOfStudy, b) the grades in Mathematics for females for the year and the region deter- mined by the values assigned to yearOfStudy and to regionOfStudy, c) the grades for males in Sciences for all regions and all years. These notation are fixed for once and will be used in similar papers. More notation related to transformed variables may come later. We post a complete R code in Appendix II in page 21 for the reading the data and forming the variable described above. In this first release of studies on GABECE, we are going to give the Gambian national context of the data and compare the representation of specific y yy yyy yyyy ´ ´ 6 OUSMANE SAINE , SOUMAILA DEMBELE , GANE SAMB LO , MOHAMED CHEIKH HAIDARA population subgroups in the data. Next we describe the main character- istics of the data under the following points: (1) facts and trends on the population; (2) incomplete data; (3) outliers; (4) general trends of the performances; THE GAMBIAN GABECE EDUCATIONAL DATA 7 CO 2012 2013 2014 2015 2016 2017 Overall Size 20850 20388 20244 21987 21943 22218 127,630 Male 10672 10378 10553 11254 11617 12032 66,506 Male (%) 51.18 50.90 52.13 51.18 52.94 54.15 52.11 Female 10178 10010 9691 10733 10326 10186 61,112 Female (%) 48.82 49.10 47.90 48.82 47.06 45.85 47.89 Region 1 6815 6717 6629 6843 6879 6832 40,715 (%) 32.69 32.95 32.75 31.12 31.35 30.75 31.90 Region 2 8182 8001 8263 9752 9519 9984 53,701 (%) 39.24 39.24 40.82 44.35 43.38 44.94 42.08 Region 3 2260 2149 2191 2090 2184 2012 12,886 (%) 10.83 10.54 10.82 9.50 9.95 9.06 10.10 Region 4 1042 1079 847 812 854 917 5,551 (%) 5.00 5.29 4.18 3.69 3.89 4.12 4.35 Region 5 1790 1603 1404 1467 1499 1388 9,151 (%) 8.59 7.86 6.94 6.67 6.83 6.25 7.17 Region 6 761 839 910 1023 1008 1085 5626 (%) 3.65 4.12 4.50 4.65 4.59 4.88 4.41 Table 1. Number of Students by year, region and gender in absolute values and in percentages. Abbreviation: CO= counted Objects. 3. In what context the GABECE data represents the whole population The data concerns young students at age around 13 passing the grade 9 exam over the period 2012-2017. The Gambia has six administrative re- gions which are : (1) Kanifing (including Banjul), (2) Brikama, (3) Kerewan, (4) Mansa Konko, (5) Janjanbureh-Kuntaur, (6) Basse. That order corre- sponds to covering the country from East to West. In particular Basse is the extrem eastern area in the west from the Capital Banjul in Kanifing. In total, 127,630 students took part to the GABECE exam and 5,792 missed at least one discipline exam or handed a blank copy. The number of declared students by year, region and gender are given in Table 1. The number of Students by year, region and gender in absolute values and in percentages are posted in table 2. y yy yyy yyyy ´ ´ 8 OUSMANE SAINE , SOUMAILA DEMBELE , GANE SAMB LO , MOHAMED CHEIKH HAIDARA CO 2012 2013 2014 2015 2016 20171 All time Region 1: Kanifing Male 3651 3627 3617 3698 3722 3856 22,171 Female 3164 3090 3012 3145 3157 2976 18,544 Male (%) 53.57 54.00 54.56 54.04 54.11 56.44 54.45 Female (%) 46.43 46.00 45.44 45.96 45.89 43.56 45.55 Region 2: Brikama Male 4002 3897 4139 4896 4968 5303 27,205 Female 4180 4104 4124 4856 4551 4681 26,469 Male (%) 48.91 48.71 50.09 50.21 52.19 53.11 50.66 Female (%) 51.09 51.29 49.91 49.79 47.81 46.89 49.34 Region 3: Mansa Konko Male 1171 1065 1161 1019 1115 1052 6,583 Female 1089 1084 1030 1071 1069 960 6,303 Male (%) 51.81 49.56 52.99 48.76 51.05 52.29 51.09 Female (%) 48.19 50.44 47.01 51.24 48.95 47.71 48.91 Region 4: Janjanbureh Male 526 516 420 384 449 483 2,778 Female 516 563 427 428 405 434 2,773 Male (%) 50.48 47.82 49.59 47.29 52.58 52.67 50.05 Female (%) 49.52 52.18 50.41 52.71 47.42 47.33 49.95 Region 5: Kerewan Male 1048 920 804 799 872 832 5,273 Female 742 683 600 668 627 556 3,876 Male (%) 58.55 57.39 57.26 54.46 58.17 59.94 57.64 Female (%) 41.45 42.61 42.74 45.54 41.83 40.06 42.36 Region 6: Basse Male 274 353 412 458 491 506 2.494 Female 487 486 498 565 517 579 3.132 Male (%) 36.01 42.07 45.27 44.77 48.71 46.64 44.33 Female (%) 63.99 57.93 54.73 55.23 51.29 53.36 55.67 Table 2. Number of Students by year, region and gender in absolute values and in percentages. CO: counted Objects. THE GAMBIAN GABECE EDUCATIONAL DATA 9 Figure 1. Evolution of number of students over the years (globally at left, for boys at the center, for girls at right) We may have expected that the number of pupils is increasing from year 2012 to 2017. However, as shown in Fig. 1, we see drops in 2013 and 2014 for the total number of students before a tendency of increase which also is broken in 2016. For boys, the positive trend is remarkable after a drop in 2013. The evolution of the number of female is somehow chaotic, suggesting a lack of a successful gender policy to get girls into the educa- tional system or to keep them in. We will compare the distribution of the regions and the gender with the national statistics. In Table 3, we have the distribution of the population over eight (8) regions (Banjul, Kanifing, Brikama, Kerewan, Kuntaur, Jan- janbureh, Basse). In Fig. 2, we have the map of the Gambia. The country, which is entirely surrounded by Sen ´ ’egal except at the Atlantic ocean coast, seems to be caught by a nearly rectangle horizontally ranging from East (Atlantic Ocean) to West in the interior of SENEGAL. In the GABECE data, the regions of Banjul and Kanifing are regrouped in one region named after the later (Kanifing) and the regions of Janjanbureh and Kauntur are regrouped in one region named after the first. In Table 4, we finally have the distribution in percentages of the population over the six geographical areas used in the GABECE data. That table also displays y yy yyy yyyy ´ ´ 10 OUSMANE SAINE , SOUMAILA DEMBELE , GANE SAMB LO , MOHAMED CHEIKH HAIDARA Figure 2. Map of the Gambia. The region of Janjanbureh and Kuntaur are regrouped and named after Kuntaur Regions Area (km2) Pop.C.03 Pop.C.13 Capital N.D Banjul (C) 12.2 35,061 31,301 Banjul 3 Kanifing 75.6 322,735 382,096 Kanifing 1 Brikama 1,764.3 389,594 699,704 Brikama 9 Mansa Konko 1,628.0 72,167 82,381 Mansakonko 6 Kerewan 2,255.5 172,835 221,054 Kerewan 7 Kuntaur 1,466.5 78,491 99,108 Kuntaur 5 Janjanbureh 1,427.8 107,212 126,910 Janjanbureh 5 Basse 2,069.5 182,586 239,916 Basse 7 Total Gambia 10,689 1,360,681 1,882,450 Banjul 43 Table 3. Population by region in 2003 and 2013 (provisional). Abbrevia- tions: pop.C.03 = population Census 2003; pop.C.13 = population Cen- sus 2013; (C) Capital of the country; N.D= number of districts. Source: Wikipedia (2020) the areas in km of each region. Fig. 3 shows the weighs of the regions in the country. We can see that the Region 1 has the greatest population density followed by the three extreme western areas. As to the gender repartition, females are slightly more dominant of males over the study period as shown in Fig. 4 THE GAMBIAN GABECE EDUCATIONAL DATA 11 Num Regions Area (km2) Pop.C.03 % Pop.C.13 % 1 Kanifing 87.4 357796 26.30% 413397 21.96% 2 Brikama 1764.3 389594 28.63% 699704 37.17% 3 Mansa Konko 1628 72167 5.30% 82381 4.38% 4 Kerewan 2255.5 172835 12.70% 221054 11.74% 5 Janjanburey 2894.3 185703 13.65% 226018 12.01% 6 Basse 2069.5 182586 13.42% 239916 12.74% Total Gambia 10689 1360681 100.00% 1882450 100.00% Table 4. Populations of regions in % in Census 2003 and census 2013 Figure 3. Populations in % of regions nationally (left) and in the data (right) y yy yyy yyyy ´ ´ 12 OUSMANE SAINE , SOUMAILA DEMBELE , GANE SAMB LO , MOHAMED CHEIKH HAIDARA Figure 4. Domination of the female gender over of male gender over years. Source: Statistica.com (2020) Now, let us see how the distribution of the GABECE compare with the na- tional data with respect to the region and the gender. a. The regions are not equally represented. The comparison of the percentages in Table 3 with those in Table 5 are summarized in Table 5. The plot of the ratios of region percentage in the data in the region percentage in the GABECE data, in Fig. 5, shows that the three western region are over represented by far in the education data, the extreme western is not almost equally represented and the three middle regions are seriously under represented. Indeed an almost equal presence of the regions in the data should have led to points very near the bisector line. The more a point is above the bisector line, the more the correspond- ing region is over represented and the more a point is below the bisector line, the more the corresponding region is under represented. b. The female gender is poorly represented in the education data. From Tables 1 and 2 and Fig. 5, it very clear that female Gambian are badly represented in the education system although they constitute the majority in the period of the study. At the exception of the Basse area, more boys arrive at Level 13 that girls from 2014 to 2018. The biggest THE GAMBIAN GABECE EDUCATIONAL DATA 13 Figure 5. Scatterplot population of regions nationally and population of regions in the data gaps happen in Kerewan. How to explain why the contrary happens in the rural zone of Basse in opposite to Kerewan area which is adjacent to it? Local studies should be y yy yyy yyyy ´ ´ 14 OUSMANE SAINE , SOUMAILA DEMBELE , GANE SAMB LO , MOHAMED CHEIKH HAIDARA Num Regions Presence in the Gambia Presence in the data 1 Kanifing 21.96% 32.95% 2 Brikama 37.17% 39.17% 3 Mansa Konko 4.38% 10.54% 4 Kerewan 11.74% 5.29% 5 Janjanburey-kuntaur 12.01% 7.83% 6 Basse 12.74% 4.14% Table 5. Population of the regions in % nationally and in the dat undergone to study such microscopic facts. In Brikama, for 2012 and 2013, girls were well enrolled in the system ac- cording to their majority but the tendency came short to be persistent. Conclusion 1. The GABECE data does not really represent the national statistics. Some regions are over represented (as the first region from East). The others are under represented with Basse being not very under present in the education data. As well, the majority of the female population is not materialized in the eduction data. This, surely, is a sign of a poor gender equity policy to be solved. Recommendation 1. In the statistical study, samplings should be used to extract from the data samples corresponding to national ratios for each region and gender to expect to have a fair comparison between of regions and of genders. 4. incomplete data The data presents a high number of incomplete data. The missing data concerns the grades G1, G2, G3, G4. For a grade, a missing data (-1 or -2) is assigned if the student did not attend the exam of that grade or he/she handed a blank copy. We only give here the percentages of missing data by region and year. We see that 2.5% to roughly 8% of the data are missing data. THE GAMBIAN GABECE EDUCATIONAL DATA 15 Regions 1 2 3 4 5 6 Data 1550 2684 519 246 378 415 Data % 3.8069 4.9980 4.0276 4.4316 4.1306 7.3764 Table 6. incomplete data by region Years 2012 2013 2014 2015 2016 2017 Data 652 589 519 1138 1709 1185 Data % 3.1270 2.8889 2.5637 5.1757 7.7883 5.3335 Table 7. incomplete data by year 5. Outliers As explained in the introduction, the grades range from very excellent (1) to fail (9), with credits between 1 to 6. Grades (7) and (8) are for passing students with very average levels. We will see later that the good students with grades (1), (2), (3) are very rare. They are so few that they seem to be outliers, that is data that can be taken as exceptional or accidental. At the same times a bulk of the grades range over f7; 8; 9g. The consequence is that outlier tests do detect students with credits as exceptional data, that should be studied apart. Here we use the simple Wilks tests for outliers which works as follows: Compute the first quartile Q , the second first quartile Q = Me 1 2 which is the median and the third quartile Q compute the interquartile distance D = Q Q 3 2 Compute the bounds d = Me 1:5D and d = Me + 1:5D. 1 2 Declare outlier at left (OL) all elements below d and outliers at right (OR) all observations beyond d We applied that simple rule to each grade G1, G2, G3 and G4. The per- centages of outliers at left (OL) and outliers at right (OR) are given by grade in Tables 8 - 11. In each table, The percentages of outliers at left (OL) and outliers at right (OR) are posted for every region. When we consider data by regions, the outliers are denoted by (OLr) [at left] and (ORr) [at left]. When data are restricted to regions and to a gender, outliers are denoted by (OLrf) [left] and (ORrf) [right] for females, (OLrm) [left] and (ORrm) [right] y yy yyy yyyy ´ ´ 16 OUSMANE SAINE , SOUMAILA DEMBELE , GANE SAMB LO , MOHAMED CHEIKH HAIDARA Regions 1 2 3 4 5 6 OLr 12.0388 14.2089 24.4279 23.9585 18.1123 20.6486 ORr 0 0 0 0 0 0 OLrf 10.6113 10.0593 11.3523 11.8554 9.1617 24.2392 ORrf 0 0 0 0 0 0 OLrm 13.2076 13.3001 20.5589 16.9184 20.2336 22.8978 ORrm 0 0 0 0 0 0 Table 8. Outliers in English grade by region and year Regions 1 2 3 4 5 6 OLr 23.4750 18.4683 16.9887 21.2629 14.9777 23.1817 ORr 0 0 0 0 0 0 OLrf 24.5746 20.3071 18.5780 9.7478 18.0010 13.9349 ORrf 0 0 0 0 0 0 OLrm 22.5746 16.7094 15.4745 17.1072 24.5693 19.3186 ORrm 0 0 0 0 0 0 Table 9. Outliers in Mathematics grades by region and year for males. The percentage of outliers range roughly between 3% and 25%. The biggest values are related to male students. The results confirm what has been said before. Conclusion 2. Excellent grades are so rare that they have been declared outliers at left. Weak grades so very common that none of them is declared outlier at right. Recommendation 2. There is no statistical interest to study the excellent grades. So it is highly recommended to regroup the grades into meang- ingful classes. For example, we may regroup them into two class: Pass (including all grades not equal to 9) and Fail for grade 9; or into three class: credit (from 1 to 6), Pass (7 and 8) and Fail (9). THE GAMBIAN GABECE EDUCATIONAL DATA 17 Regions 1 2 3 4 5 6 OLr 14.3138 10.7826 23.8861 10.2356 18.4429 18.6336 ORr 0 0 0 0 0 0 OLrf 15.0862 12.6854 10.8883 14.1512 9.0053 22.3374 ORrf 0 0 0 0 0 0 OLrm 13.6813 23.3288 19.7852 18.2779 23.2627 21.5179 ORrm 0 0 0 0 0 0 Table 10. Outliers in Sciences grades by region and year Regions 1 2 3 4 5 6 OLr 4.6572 14.1776 12.0805 2.7144 12.1395 12.9533 ORr 0 0 0 0 0 0 OLrf 0 3.4279 2.9830 0 2.5523 13.0013 ORrf 0 0 0 0 0 0 OLrm 4.4489 14.9146 11.6374 11.8957 23.0647 24.5364 ORrm 0 0 0 0 0 0 Table 11. Outliers in SES grades by region and year 6. General analysis of performances We are giving some trends of the performances. More insight studies will be made in up coming papers. Table 12 gives the percentages of grades (1 to 9) by discipline and year. We can see the following conclusions: (C1) The fail grade is overwhelmingly dominant in Mathematics, Sciences and English disciplines. Even for SES which can be thought less tough that the three others, fail rates are significant and turn around 50%. (C2) The excellency credit grade (1) is extremely rare and its rate does not exceed 2% at the exception of the disciplines SES for which it is still low and less than 5%. (C3) The other credit grades are still low enough to be studied apart. y yy yyy yyyy ´ ´ 18 OUSMANE SAINE , SOUMAILA DEMBELE , GANE SAMB LO , MOHAMED CHEIKH HAIDARA disp/year cr 1 cr 2 cr 3 ce 4 cr 5 cr 6 pass 1 pass 2 fail G1A 2012 1.75 2.13 6.62 4.44 3.73 8.76 6.94 9.66 55.97 G2A 2012 0.54 0.72 2.20 1.59 1.82 5.69 8.12 12.99 66.33 G3A 2012 0.43 0.60 2.91 2.19 3.00 7.85 4.94 7.88 70.19 G4A 2012 3.10 2.89 9.52 4.41 5.15 12.60 5.50 10.57 46.27 G1A 2013 0.56 0.81 3.23 2.66 2.45 6.72 7.61 10.60 65.35 G2A 2013 0.67 1.03 3.73 2.34 2.76 7.17 6.84 10.14 65.31 G3A 2013 0.31 0.53 2.46 1.75 1.74 8.33 6.65 11.24 66.98 G4A 2013 2.10 2.48 8.16 3.89 4.65 11.30 7.09 10.20 50.12 G1A 2014 0.19 0.60 2.93 2.48 2.34 8.97 9.67 12.48 60.35 G2A 2014 0.60 0.61 2.15 0.76 1.53 5.28 6.02 10.33 72.71 G3A 2014 1.80 1.12 3.90 2.62 2.40 8.15 8.61 14.68 56.74 G4A 2014 2.50 2.54 9.22 4.55 5.11 12.29 5.42 10.40 47.96 G1A 2015 1.27 1.53 6.27 4.06 3.93 11.14 9.69 11.37 50.74 G2A 2015 0.74 0.69 2.59 1.39 1.82 5.21 2.49 15.83 69.25 G3A 2015 0.29 0.58 2.83 1.32 2.36 8.07 9.59 14.60 60.38 G4A 2015 2.53 2.46 7.18 4.93 3.38 11.46 7.46 10.87 49.71 G1A 2016 0.32 1.00 5.90 3.18 5.33 10.07 7.94 11.08 55.17 G1A 2016 0.85 0.86 2.98 1.50 1.48 5.77 7.15 13.70 65.69 G2A 2016 1.45 1.36 4.19 2.90 2.18 8.50 8.01 10.90 60.51 G3A 2016 3.18 3.56 9.39 5.36 5.27 13.48 6.06 8.68 45.03 G4A 2017 0.45 1.11 6.13 4.21 10.34 7.59 8.00 9.14 53.03 G2A 2017 0.69 0.82 2.68 1.14 4.04 3.54 6.71 12.74 67.64 G3A 2017 1.82 1.56 5.21 2.90 7.69 5.92 7.54 8.24 59.12 G4A 2017 4.88 4.03 10.91 4.93 11.40 7.16 7.07 7.74 41.88 Table 12. Percentages of grades (1 to 9) by disciplines and by year 7. Conclusions This paper is destined to be the basis of a series of more insightful papers using a variety of more or less deep statistical techniques. At least, the data is entirely described. The notations are posed. A general algorithm allowing to get all relevant variables and their local restrictions (in time (year), in area (region), in gender (female, male)). First statistical trends are given, concerning incomplete data, outliers, fair- ness of representation of sub-groups in the educational data, some general evolutions of grades in the time. These statistical facts already direct to THE GAMBIAN GABECE EDUCATIONAL DATA 19 garde/year cr1-4 cr1-5 cr1-6 pass fail G1A 2012 14.94 18.67 27.43 44.03 55.97 G2A 2012 5.05 6.87 12.56 33.67 66.33 G3A 2012 6.13 9.13 16.98 29.8 70.19 G4A 2012 19.92 25.07 37.67 53.74 46.27 G1A 2013 7.26 9.71 16.43 34.64 65.35 G2A 2013 7.77 10.53 17.7 34.68 65.31 G3A 2013 5.05 6.79 15.12 33.01 66.98 G4A 2013 16.63 21.28 32.58 49.87 50.12 G1A 2014 6.2 8.54 17.51 39.66 60.35 G2A 2014 4.12 5.65 10.93 27.28 72.71 G3A 2014 9.44 11.84 19.99 43.28 56.74 G4A 2014 18.81 23.92 36.21 52.03 47.96 G1A 2015 13.13 17.06 28.2 49.26 50.74 G2A 2015 5.41 7.23 12.44 30.76 69.25 G3A 2015 5.02 7.38 15.45 39.64 60.38 G4A 2015 17.1 20.48 31.94 50.27 49.71 G1A 2016 10.4 15.73 25.8 44.82 55.17 G1A 2016 6.19 7.67 13.44 34.29 65.69 G2A 2016 9.9 12.08 20.58 39.49 60.51 G3A 2016 21.49 26.76 40.24 54.98 45.03 G4A 2017 11.9 22.24 29.83 46.97 53.03 G2A 2017 5.33 9.37 12.91 32.36 67.64 G3A 2017 11.49 19.18 25.1 40.88 59.12 G4A 2017 24.75 36.15 43.31 58.12 41.88 Table 13. Percentages of regrouped grades (credit, pass, fail) by discipline and by year interesting leads and suggest hypotheses to be validated or not. Acknowledgement The authors wish to thank the Ministry .... that al- lowed to use the data. References Crawley M. J.(2009). The R book. Wiley. James G., Witten D., Hastie T. and Tibshirani R.(2008) Introduction to th statistical learning with Applications in R. (Corrected at 8 edition 2017). Springer y yy yyy yyyy ´ ´ 20 OUSMANE SAINE , SOUMAILA DEMBELE , GANE SAMB LO , MOHAMED CHEIKH HAIDARA R Core Team (2020). R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria. https://www.R-project.org. Statistica com. https://www.statista.com/statistics/967843/total- population-of-gambia-by-gender/. Visited on 2020/11/13. Wikipedia website. https://en.wikipedia.org/wiki/The Gambia#Administrative divisions. Visited on 2020/11/11. THE GAMBIAN GABECE EDUCATIONAL DATA 21 Appendix : R codes See Crawley (2009) for an overall review on R software. #The data are i n the d i r e c t o r y gabece , which i s # o f a s u b d i r e c t o r y where i s t h i s s c r i p t . #The d i r e c t o r y should contain the f o l l o w i n g f i l e s r e l a t i o n t o #the whole population o f students , having the same lenghth #1 − mesAppurements . t x t : contains the v a r i a b l e okData with two values #”ok ” f o r complete data ” ko ” f o r incomplete data .nn #2 − the o r i g i n a l grades G1, G2, G2, and G1. #3 − the c o v a r i a t e s year . t x t , gender . t x t and r e g i o n . t x t appu <− read . t a b l e ( ” gabece/mesAppurements . t x t ” , h=TRUE) attach ( appu ) t a b l e ( okData ) Gender <− read . t a b l e ( ” gabece/gender . t x t ” , h=TRUE) attach ( Gender ) gender t a b l e ( gender ) Year <− read . t a b l e ( ” gabece/year . t x t ” , h=TRUE) attach ( Year ) t a b l e ( year ) Region <− read . t a b l e ( ” gabece/ r e g i o n . t x t ” , h=TRUE) attach ( Region ) t a b l e ( r e g i o n ) English <− read . t a b l e ( ” gabece/G1. t x t ” , h=TRUE) attach ( English ) t a b l e ( G1 ) Maths <− read . t a b l e ( ” gabece/G2. t x t ” , h=TRUE) attach ( Maths ) t a b l e ( G2 ) Sciences <− read . t a b l e ( ” gabece/G3. t x t ” , h=TRUE) attach ( Sciences ) y yy yyy yyyy ´ ´ 22 OUSMANE SAINE , SOUMAILA DEMBELE , GANE SAMB LO , MOHAMED CHEIKH HAIDARA t a b l e ( G3 ) SES <− read . t a b l e ( ” gabece/G4. t x t ” , h=TRUE) attach ( SES ) t a b l e ( G4 ) # Cleaning ============================================= # V a r i a b l e s i n c l u d i n g cases not p r e s e n t i n g missing data −1 , −2 , e t c . . genderA=gender [ okData==”ok ” ] yearA=year [ okData==”ok ” ] regionA= r e g i o n [ okData==”ok ” ] G1A=G1 [ okData==”ok ” ] G2A=G2 [ okData==”ok ” ] G3A=G3 [ okData==”ok ” ] G4A=G4 [ okData==”ok ” ] # ======================================== # y e a r l y A n a l y s i s #In the example below , you may change 2012 #by any year i n (2012 , . . . , 2 0 1 7 ) and #proceed t o the a n a l y s i s yearOfStudy=2012 regionOfStudy = ” 6 ” genderAY=genderA [ yearA==yearOfStudy ] regionAY=regionA [ yearA==yearOfStudy ] G1AY=G1A[ yearA==yearOfStudy ] G2AY=G2A[ yearA==yearOfStudy ] G3AY=G3A[ yearA==yearOfStudy ] G4AY=G4A[ yearA==yearOfStudy ] # study o f r e g i o n genderAR=genderA [ regionA==regionOfStudy ] G1AR=G1A[ regionA==regionOfStudy ] G2AR=G2A[ regionA==regionOfStudy ] G3AR=G3A[ regionA==regionOfStudy ] G4AR=G4A[ regionA==regionOfStudy ] # study o f r e g i o n and year THE GAMBIAN GABECE EDUCATIONAL DATA 23 genderAYR=genderA [ ( regionA==regionOfStudy ) & ( yearA==yearOfStudy ) ] G1AYR=G1A [ ( regionA==regionOfStudy ) & ( yearA==yearOfStudy ) ] G2AYR=G2A [ ( regionA==regionOfStudy ) & ( yearA==yearOfStudy ) ] G3AYR=G3A [ ( regionA==regionOfStudy ) & ( yearA==yearOfStudy ) ] G4AYR=G4A [ ( regionA==regionOfStudy ) & ( yearA==yearOfStudy ) ] l e n g t h (G4AYR) #proceed t o your a n a l y s i s http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Statistics arXiv (Cornell University)

Overview description of the Gambian GABECE Educational Data and associated algorithms and unsupervized learning process

Download PDF
23 pages

Loading next page...
 
/lp/arxiv-cornell-university/overview-description-of-the-gambian-gabece-educational-data-and-Y5H9MMOU4t
ISSN
2316-0861
eISSN
ARCH-3347
DOI
10.16929/ajas/2020.451.245
Publisher site
See Article on Publisher Site

Abstract

OVERVIEW DESCRIPTION OF THE GAMBIAN GABECE EDUCATIONAL DATA AND ASSOCIATED ALGORITHMS AND UNSUPERVIZED LEARNING PROCESS y yy yyy yyyy ´ ´ OUSMANE SAINE , SOUMAILA DEMBELE , GANE SAMB LO , MOHAMED CHEIKH HAIDARA Abstract. As the first paper of a series of exploratory analysis and statistical investigation works on the Gambian GABECE data based on a variety of statistical tools, we wish to begin with a thorough unsupervised learning process through descriptive and exploratory meth- ods. This will lead to a variety of discoveries and hypotheses that will direct future research works related to this data. Ousman Saine University of The Gambia, Banjul, The Gambia & Policy Planning, Analyses, research and Budgeting Directorate, Ministry of Basic and secondary Education, Willy Thorpe Palace, Banjul, The Gambia ousmansaine@edugambia.gm yy Dr Soumaila DEMBELE Universite ´ des Sciences Sociale et de Gestion de Bamako ( USSGB) Faculte ´ des Sciences Economiques et de Gestion (FSEG) Email: soumidemlpot@gmail.com yyy Gane Samb Lo. LERSTAD, Gaston Berger University, Saint-Louis, Sen ´ eg ´ al (main affiliation). LSTA, Pierre and Marie Curie University, Paris VI, France. AUST - African University of Sciences and Technology, Abuja, Nigeria gane-samb.lo@edu.ugb.sn, gslo@aust.edu.ng, ganesamblo@ganesamblo.net Permanent address : 1178 Evanston Dr NW T3P 0J9,Calgary, Alberta, Canada. yyyy Dr Mohamed Cheikh HAIDARA Work Affiliation : Depar ´ tement de Mathematiq ´ ue , Universite ´ Cheikh Anta Diop (UCAD), Dakar, Senegal Ecole Super ´ ieure Polytechnique, Universite ´ Cheikh Anta Diop (UCAD), Dakar, Senegal Research affiliation : LERSTAD, Universite ´ Gaston Berger (UGB), Saint-Louis, Senegal Emails : chheikhh@yahoo.fr Keywords. record values; record time; endpoints of a cumulative distribution function; hit- ting times; central limit theorem, law of the iterated logarithm; Berry-Essen bound; band of consistency; confidence intervals. AMS 2010 Mathematics Subject Classification: 62G30; 60Fxx; 63H10 arXiv:2012.08488v1 [stat.AP] 15 Dec 2020 y yy yyy yyyy ´ ´ 2 OUSMANE SAINE , SOUMAILA DEMBELE , GANE SAMB LO , MOHAMED CHEIKH HAIDARA 1. Introduction In this paper, the GABECE data is first presented and next statistically studied for the first time. We proceed to an unsupervised learning pro- cess. We will provide the description of the variables and consider the geographical and temporal sub-data. For each sub-data, different tech- niques will be applied : cleaning, outliers detection, frequency and related tables, spatial (geographical) or gender patterns, proportions of grades and evolution in time, etc. A series of statistical analyzes will be undergone on the data by ourselves or other researchers. To prepare that, we wish to devote a note on the gen- eral presentation of the data and an unsupervised learning process with respect to the gender, the area and the time. Also R codes (R Core Team (2020) and Crawley (2009)) aiming at reading the data and getting the variables presented in Section 2 will be provided in Appendix II in page 21. The rest of the paper is organized as follows. We will complete this intro- duction by the main findings in this paper and related recommendations. The data is fully described in Section 2. In Section 3, the GABECE data is compared with national statistics concerning population sizes of regions and genders to see whether or not access to education is fair in the Gambia. In section 4, the completeness of the the data is studied and it it will be shown that the data present un non-negligible number of incomplete data everywhere. In Section 5, a cleaning study is undergone by the detection of the outliers and it will appear that eliminating outliers, as usually rec- ommended, would lead to a serious error since excellent students would simply ignored in the study. Instead of rejecting, we will propose regroup- ing the grades. In Section 6, we undergo a general investigation of the performances. We close the paper by conclusions in Section 7. Appen- dices are put after the bibliography. . The main conclusions are summarized below. (C1) Neither the regions nor the two genders are well represented in the educational GABECE data. The western regions, in general, have more students at grade 13 when compared with these real weighs in the coun- try. Although being more numerous in the country, females are under repre- sented in the data. Even worse, the number of females in the data behaves THE GAMBIAN GABECE EDUCATIONAL DATA 3 chaotically where the number of males steadily increases with the year. (R1) As a consequence, we strongly recommend to use sampling meth- ods with sample sizes proportional to the real weighs of sub-groups in the global population. (C2) There is a big number of incomplete data in all years, in all regions, and in each gender. After dropping them, we still have enough data to proceed to a significant analysis data. (R2) On this side, we recommend statistical studies based on missing data estimation techniques. There is a lot to do on this, using nearest neigh- borhood, regression, classification, etc. (C3) The statistical studies are about the grades in the core disciplines: Mathematics, Sciences, English, and SES. The grades range from 1 (ex- cellent) to 9 (fail). Unfortunately, the level of students is not very fairly good since the fail grade is dominant almost everywhere. As well, there is a very few number of excellent students. (R3) The result of that situation is that excellent grades 1, 2 or 3 are de- tected as outliers. So, in the statistical studies, such grades should be excluded. But, fortunately, a regrouping of all grades into three levels seems to be the best way to overcome that problem. Those three levels are: credit (regrouping credit grades 1 to 6), pass (regrouping 8 and 8) and fail. The paper is destined to be a basis for a series of statistical studies on GABECE data and beyond on similar data for countries of the West African Examinations Council, whose The Gambia is member of, and on educa- tional data in general. 2. Overall presentation of the GABECE data The West African Examinations Council was established in 1952. This Council has the responsibility of ascertaining examinations required in the public interest in the English-speaking West African countries, com- parable to those of equivalent examining authorities internationally, and to develop syllabuses, arranges and administers examinations and awards y yy yyy yyyy ´ ´ 4 OUSMANE SAINE , SOUMAILA DEMBELE , GANE SAMB LO , MOHAMED CHEIKH HAIDARA certificates. It also conducts a number international examinations. The data we are going to use in this document is from the above afore- mentioned body. The data covers the period from the 2012 academic year to 2017. The mode of the GABECE examination is that each student is required to choose 7 subjects at least and 9 subjects at most. Out of the subjects taken, four (4) are compulsory and are called core subjects, which are English (ENG), Mathematics (MATH), Sciences (SCIENCES) and Social and Environmental Studies (SES). Individual performance grants the student admission to Senior Secondary Education. The grading system in this result is categorized as follows : Credit (grades from 1 to 6), Pass (7 to 8), Fail (9). The result of a student is determined by the grades of the four core subjects plus the two best grades from the student’s choice. The best result a student could have is the aggregated grade from 6 (six), one in all the core subjects, plus one in the two choices of the students and the worst result is aggregated to 54 with failure in all subjects with grade 9. In this study, we focus on the core grades (G1), (G2), (G3) and (G4) corre- sponding to English, Mathematics, Sciences and SES respectively. Here are the conventions we will be using to know whether we use a grade for the whole population or not, for all the years or not, for both genders or not, or whether we use a grade for one particular gender, or a year or an area. The corresponding variables of the core grades are denoted as G1, G2, G3 and G4 before the cleaning of the data and cover all areas, all time and all gender. We also have the variables year (from 2012 to 2017), gender (coded 1 for males and 2 for females), and region (coded for 1 to six regions). The missing data, which correspond to absences of students or submis- sion of blank sheets by student, represent 4.54% of the data (5792 out of 121838). The new variables become : G1A, G2A, G3A, G4A, yearA, genderA, regionA. THE GAMBIAN GABECE EDUCATIONAL DATA 5 When we restrict our analysis on a specific year for all gender, the variables are: G1AY, G2AY, G3AY, G4AY, yearAY, genderAY, regionAY. When we restrict our analysis on a specific region for all genders and years, the variables are : G1AR, G2AR, G3AR, G4AR, yearAR, genderAR, regionAR. When we restrict our analysis on a specific region and for a specific year, for all genders, the variables are : G1AYR, G2AYR, G3AYR, G4AYR, yearAYR, genderAYR, regionAYR. In each of the previous cases, the specified year and region are given by the variables yearOfStudy and regionOfStudy. If we want to specify and fix a gender, we use the filtering codes : Vari- able[gender==”1”] for males (and Variable[gender==”2”] for females). For example, the variables G1AR[genderAR==”1”], G2AYR[genderAYR==”2”], G3A[genderA==”1”], respectively contain: a) the grades for males in English for the region determined by the value of regionOfStudy, b) the grades in Mathematics for females for the year and the region deter- mined by the values assigned to yearOfStudy and to regionOfStudy, c) the grades for males in Sciences for all regions and all years. These notation are fixed for once and will be used in similar papers. More notation related to transformed variables may come later. We post a complete R code in Appendix II in page 21 for the reading the data and forming the variable described above. In this first release of studies on GABECE, we are going to give the Gambian national context of the data and compare the representation of specific y yy yyy yyyy ´ ´ 6 OUSMANE SAINE , SOUMAILA DEMBELE , GANE SAMB LO , MOHAMED CHEIKH HAIDARA population subgroups in the data. Next we describe the main character- istics of the data under the following points: (1) facts and trends on the population; (2) incomplete data; (3) outliers; (4) general trends of the performances; THE GAMBIAN GABECE EDUCATIONAL DATA 7 CO 2012 2013 2014 2015 2016 2017 Overall Size 20850 20388 20244 21987 21943 22218 127,630 Male 10672 10378 10553 11254 11617 12032 66,506 Male (%) 51.18 50.90 52.13 51.18 52.94 54.15 52.11 Female 10178 10010 9691 10733 10326 10186 61,112 Female (%) 48.82 49.10 47.90 48.82 47.06 45.85 47.89 Region 1 6815 6717 6629 6843 6879 6832 40,715 (%) 32.69 32.95 32.75 31.12 31.35 30.75 31.90 Region 2 8182 8001 8263 9752 9519 9984 53,701 (%) 39.24 39.24 40.82 44.35 43.38 44.94 42.08 Region 3 2260 2149 2191 2090 2184 2012 12,886 (%) 10.83 10.54 10.82 9.50 9.95 9.06 10.10 Region 4 1042 1079 847 812 854 917 5,551 (%) 5.00 5.29 4.18 3.69 3.89 4.12 4.35 Region 5 1790 1603 1404 1467 1499 1388 9,151 (%) 8.59 7.86 6.94 6.67 6.83 6.25 7.17 Region 6 761 839 910 1023 1008 1085 5626 (%) 3.65 4.12 4.50 4.65 4.59 4.88 4.41 Table 1. Number of Students by year, region and gender in absolute values and in percentages. Abbreviation: CO= counted Objects. 3. In what context the GABECE data represents the whole population The data concerns young students at age around 13 passing the grade 9 exam over the period 2012-2017. The Gambia has six administrative re- gions which are : (1) Kanifing (including Banjul), (2) Brikama, (3) Kerewan, (4) Mansa Konko, (5) Janjanbureh-Kuntaur, (6) Basse. That order corre- sponds to covering the country from East to West. In particular Basse is the extrem eastern area in the west from the Capital Banjul in Kanifing. In total, 127,630 students took part to the GABECE exam and 5,792 missed at least one discipline exam or handed a blank copy. The number of declared students by year, region and gender are given in Table 1. The number of Students by year, region and gender in absolute values and in percentages are posted in table 2. y yy yyy yyyy ´ ´ 8 OUSMANE SAINE , SOUMAILA DEMBELE , GANE SAMB LO , MOHAMED CHEIKH HAIDARA CO 2012 2013 2014 2015 2016 20171 All time Region 1: Kanifing Male 3651 3627 3617 3698 3722 3856 22,171 Female 3164 3090 3012 3145 3157 2976 18,544 Male (%) 53.57 54.00 54.56 54.04 54.11 56.44 54.45 Female (%) 46.43 46.00 45.44 45.96 45.89 43.56 45.55 Region 2: Brikama Male 4002 3897 4139 4896 4968 5303 27,205 Female 4180 4104 4124 4856 4551 4681 26,469 Male (%) 48.91 48.71 50.09 50.21 52.19 53.11 50.66 Female (%) 51.09 51.29 49.91 49.79 47.81 46.89 49.34 Region 3: Mansa Konko Male 1171 1065 1161 1019 1115 1052 6,583 Female 1089 1084 1030 1071 1069 960 6,303 Male (%) 51.81 49.56 52.99 48.76 51.05 52.29 51.09 Female (%) 48.19 50.44 47.01 51.24 48.95 47.71 48.91 Region 4: Janjanbureh Male 526 516 420 384 449 483 2,778 Female 516 563 427 428 405 434 2,773 Male (%) 50.48 47.82 49.59 47.29 52.58 52.67 50.05 Female (%) 49.52 52.18 50.41 52.71 47.42 47.33 49.95 Region 5: Kerewan Male 1048 920 804 799 872 832 5,273 Female 742 683 600 668 627 556 3,876 Male (%) 58.55 57.39 57.26 54.46 58.17 59.94 57.64 Female (%) 41.45 42.61 42.74 45.54 41.83 40.06 42.36 Region 6: Basse Male 274 353 412 458 491 506 2.494 Female 487 486 498 565 517 579 3.132 Male (%) 36.01 42.07 45.27 44.77 48.71 46.64 44.33 Female (%) 63.99 57.93 54.73 55.23 51.29 53.36 55.67 Table 2. Number of Students by year, region and gender in absolute values and in percentages. CO: counted Objects. THE GAMBIAN GABECE EDUCATIONAL DATA 9 Figure 1. Evolution of number of students over the years (globally at left, for boys at the center, for girls at right) We may have expected that the number of pupils is increasing from year 2012 to 2017. However, as shown in Fig. 1, we see drops in 2013 and 2014 for the total number of students before a tendency of increase which also is broken in 2016. For boys, the positive trend is remarkable after a drop in 2013. The evolution of the number of female is somehow chaotic, suggesting a lack of a successful gender policy to get girls into the educa- tional system or to keep them in. We will compare the distribution of the regions and the gender with the national statistics. In Table 3, we have the distribution of the population over eight (8) regions (Banjul, Kanifing, Brikama, Kerewan, Kuntaur, Jan- janbureh, Basse). In Fig. 2, we have the map of the Gambia. The country, which is entirely surrounded by Sen ´ ’egal except at the Atlantic ocean coast, seems to be caught by a nearly rectangle horizontally ranging from East (Atlantic Ocean) to West in the interior of SENEGAL. In the GABECE data, the regions of Banjul and Kanifing are regrouped in one region named after the later (Kanifing) and the regions of Janjanbureh and Kauntur are regrouped in one region named after the first. In Table 4, we finally have the distribution in percentages of the population over the six geographical areas used in the GABECE data. That table also displays y yy yyy yyyy ´ ´ 10 OUSMANE SAINE , SOUMAILA DEMBELE , GANE SAMB LO , MOHAMED CHEIKH HAIDARA Figure 2. Map of the Gambia. The region of Janjanbureh and Kuntaur are regrouped and named after Kuntaur Regions Area (km2) Pop.C.03 Pop.C.13 Capital N.D Banjul (C) 12.2 35,061 31,301 Banjul 3 Kanifing 75.6 322,735 382,096 Kanifing 1 Brikama 1,764.3 389,594 699,704 Brikama 9 Mansa Konko 1,628.0 72,167 82,381 Mansakonko 6 Kerewan 2,255.5 172,835 221,054 Kerewan 7 Kuntaur 1,466.5 78,491 99,108 Kuntaur 5 Janjanbureh 1,427.8 107,212 126,910 Janjanbureh 5 Basse 2,069.5 182,586 239,916 Basse 7 Total Gambia 10,689 1,360,681 1,882,450 Banjul 43 Table 3. Population by region in 2003 and 2013 (provisional). Abbrevia- tions: pop.C.03 = population Census 2003; pop.C.13 = population Cen- sus 2013; (C) Capital of the country; N.D= number of districts. Source: Wikipedia (2020) the areas in km of each region. Fig. 3 shows the weighs of the regions in the country. We can see that the Region 1 has the greatest population density followed by the three extreme western areas. As to the gender repartition, females are slightly more dominant of males over the study period as shown in Fig. 4 THE GAMBIAN GABECE EDUCATIONAL DATA 11 Num Regions Area (km2) Pop.C.03 % Pop.C.13 % 1 Kanifing 87.4 357796 26.30% 413397 21.96% 2 Brikama 1764.3 389594 28.63% 699704 37.17% 3 Mansa Konko 1628 72167 5.30% 82381 4.38% 4 Kerewan 2255.5 172835 12.70% 221054 11.74% 5 Janjanburey 2894.3 185703 13.65% 226018 12.01% 6 Basse 2069.5 182586 13.42% 239916 12.74% Total Gambia 10689 1360681 100.00% 1882450 100.00% Table 4. Populations of regions in % in Census 2003 and census 2013 Figure 3. Populations in % of regions nationally (left) and in the data (right) y yy yyy yyyy ´ ´ 12 OUSMANE SAINE , SOUMAILA DEMBELE , GANE SAMB LO , MOHAMED CHEIKH HAIDARA Figure 4. Domination of the female gender over of male gender over years. Source: Statistica.com (2020) Now, let us see how the distribution of the GABECE compare with the na- tional data with respect to the region and the gender. a. The regions are not equally represented. The comparison of the percentages in Table 3 with those in Table 5 are summarized in Table 5. The plot of the ratios of region percentage in the data in the region percentage in the GABECE data, in Fig. 5, shows that the three western region are over represented by far in the education data, the extreme western is not almost equally represented and the three middle regions are seriously under represented. Indeed an almost equal presence of the regions in the data should have led to points very near the bisector line. The more a point is above the bisector line, the more the correspond- ing region is over represented and the more a point is below the bisector line, the more the corresponding region is under represented. b. The female gender is poorly represented in the education data. From Tables 1 and 2 and Fig. 5, it very clear that female Gambian are badly represented in the education system although they constitute the majority in the period of the study. At the exception of the Basse area, more boys arrive at Level 13 that girls from 2014 to 2018. The biggest THE GAMBIAN GABECE EDUCATIONAL DATA 13 Figure 5. Scatterplot population of regions nationally and population of regions in the data gaps happen in Kerewan. How to explain why the contrary happens in the rural zone of Basse in opposite to Kerewan area which is adjacent to it? Local studies should be y yy yyy yyyy ´ ´ 14 OUSMANE SAINE , SOUMAILA DEMBELE , GANE SAMB LO , MOHAMED CHEIKH HAIDARA Num Regions Presence in the Gambia Presence in the data 1 Kanifing 21.96% 32.95% 2 Brikama 37.17% 39.17% 3 Mansa Konko 4.38% 10.54% 4 Kerewan 11.74% 5.29% 5 Janjanburey-kuntaur 12.01% 7.83% 6 Basse 12.74% 4.14% Table 5. Population of the regions in % nationally and in the dat undergone to study such microscopic facts. In Brikama, for 2012 and 2013, girls were well enrolled in the system ac- cording to their majority but the tendency came short to be persistent. Conclusion 1. The GABECE data does not really represent the national statistics. Some regions are over represented (as the first region from East). The others are under represented with Basse being not very under present in the education data. As well, the majority of the female population is not materialized in the eduction data. This, surely, is a sign of a poor gender equity policy to be solved. Recommendation 1. In the statistical study, samplings should be used to extract from the data samples corresponding to national ratios for each region and gender to expect to have a fair comparison between of regions and of genders. 4. incomplete data The data presents a high number of incomplete data. The missing data concerns the grades G1, G2, G3, G4. For a grade, a missing data (-1 or -2) is assigned if the student did not attend the exam of that grade or he/she handed a blank copy. We only give here the percentages of missing data by region and year. We see that 2.5% to roughly 8% of the data are missing data. THE GAMBIAN GABECE EDUCATIONAL DATA 15 Regions 1 2 3 4 5 6 Data 1550 2684 519 246 378 415 Data % 3.8069 4.9980 4.0276 4.4316 4.1306 7.3764 Table 6. incomplete data by region Years 2012 2013 2014 2015 2016 2017 Data 652 589 519 1138 1709 1185 Data % 3.1270 2.8889 2.5637 5.1757 7.7883 5.3335 Table 7. incomplete data by year 5. Outliers As explained in the introduction, the grades range from very excellent (1) to fail (9), with credits between 1 to 6. Grades (7) and (8) are for passing students with very average levels. We will see later that the good students with grades (1), (2), (3) are very rare. They are so few that they seem to be outliers, that is data that can be taken as exceptional or accidental. At the same times a bulk of the grades range over f7; 8; 9g. The consequence is that outlier tests do detect students with credits as exceptional data, that should be studied apart. Here we use the simple Wilks tests for outliers which works as follows: Compute the first quartile Q , the second first quartile Q = Me 1 2 which is the median and the third quartile Q compute the interquartile distance D = Q Q 3 2 Compute the bounds d = Me 1:5D and d = Me + 1:5D. 1 2 Declare outlier at left (OL) all elements below d and outliers at right (OR) all observations beyond d We applied that simple rule to each grade G1, G2, G3 and G4. The per- centages of outliers at left (OL) and outliers at right (OR) are given by grade in Tables 8 - 11. In each table, The percentages of outliers at left (OL) and outliers at right (OR) are posted for every region. When we consider data by regions, the outliers are denoted by (OLr) [at left] and (ORr) [at left]. When data are restricted to regions and to a gender, outliers are denoted by (OLrf) [left] and (ORrf) [right] for females, (OLrm) [left] and (ORrm) [right] y yy yyy yyyy ´ ´ 16 OUSMANE SAINE , SOUMAILA DEMBELE , GANE SAMB LO , MOHAMED CHEIKH HAIDARA Regions 1 2 3 4 5 6 OLr 12.0388 14.2089 24.4279 23.9585 18.1123 20.6486 ORr 0 0 0 0 0 0 OLrf 10.6113 10.0593 11.3523 11.8554 9.1617 24.2392 ORrf 0 0 0 0 0 0 OLrm 13.2076 13.3001 20.5589 16.9184 20.2336 22.8978 ORrm 0 0 0 0 0 0 Table 8. Outliers in English grade by region and year Regions 1 2 3 4 5 6 OLr 23.4750 18.4683 16.9887 21.2629 14.9777 23.1817 ORr 0 0 0 0 0 0 OLrf 24.5746 20.3071 18.5780 9.7478 18.0010 13.9349 ORrf 0 0 0 0 0 0 OLrm 22.5746 16.7094 15.4745 17.1072 24.5693 19.3186 ORrm 0 0 0 0 0 0 Table 9. Outliers in Mathematics grades by region and year for males. The percentage of outliers range roughly between 3% and 25%. The biggest values are related to male students. The results confirm what has been said before. Conclusion 2. Excellent grades are so rare that they have been declared outliers at left. Weak grades so very common that none of them is declared outlier at right. Recommendation 2. There is no statistical interest to study the excellent grades. So it is highly recommended to regroup the grades into meang- ingful classes. For example, we may regroup them into two class: Pass (including all grades not equal to 9) and Fail for grade 9; or into three class: credit (from 1 to 6), Pass (7 and 8) and Fail (9). THE GAMBIAN GABECE EDUCATIONAL DATA 17 Regions 1 2 3 4 5 6 OLr 14.3138 10.7826 23.8861 10.2356 18.4429 18.6336 ORr 0 0 0 0 0 0 OLrf 15.0862 12.6854 10.8883 14.1512 9.0053 22.3374 ORrf 0 0 0 0 0 0 OLrm 13.6813 23.3288 19.7852 18.2779 23.2627 21.5179 ORrm 0 0 0 0 0 0 Table 10. Outliers in Sciences grades by region and year Regions 1 2 3 4 5 6 OLr 4.6572 14.1776 12.0805 2.7144 12.1395 12.9533 ORr 0 0 0 0 0 0 OLrf 0 3.4279 2.9830 0 2.5523 13.0013 ORrf 0 0 0 0 0 0 OLrm 4.4489 14.9146 11.6374 11.8957 23.0647 24.5364 ORrm 0 0 0 0 0 0 Table 11. Outliers in SES grades by region and year 6. General analysis of performances We are giving some trends of the performances. More insight studies will be made in up coming papers. Table 12 gives the percentages of grades (1 to 9) by discipline and year. We can see the following conclusions: (C1) The fail grade is overwhelmingly dominant in Mathematics, Sciences and English disciplines. Even for SES which can be thought less tough that the three others, fail rates are significant and turn around 50%. (C2) The excellency credit grade (1) is extremely rare and its rate does not exceed 2% at the exception of the disciplines SES for which it is still low and less than 5%. (C3) The other credit grades are still low enough to be studied apart. y yy yyy yyyy ´ ´ 18 OUSMANE SAINE , SOUMAILA DEMBELE , GANE SAMB LO , MOHAMED CHEIKH HAIDARA disp/year cr 1 cr 2 cr 3 ce 4 cr 5 cr 6 pass 1 pass 2 fail G1A 2012 1.75 2.13 6.62 4.44 3.73 8.76 6.94 9.66 55.97 G2A 2012 0.54 0.72 2.20 1.59 1.82 5.69 8.12 12.99 66.33 G3A 2012 0.43 0.60 2.91 2.19 3.00 7.85 4.94 7.88 70.19 G4A 2012 3.10 2.89 9.52 4.41 5.15 12.60 5.50 10.57 46.27 G1A 2013 0.56 0.81 3.23 2.66 2.45 6.72 7.61 10.60 65.35 G2A 2013 0.67 1.03 3.73 2.34 2.76 7.17 6.84 10.14 65.31 G3A 2013 0.31 0.53 2.46 1.75 1.74 8.33 6.65 11.24 66.98 G4A 2013 2.10 2.48 8.16 3.89 4.65 11.30 7.09 10.20 50.12 G1A 2014 0.19 0.60 2.93 2.48 2.34 8.97 9.67 12.48 60.35 G2A 2014 0.60 0.61 2.15 0.76 1.53 5.28 6.02 10.33 72.71 G3A 2014 1.80 1.12 3.90 2.62 2.40 8.15 8.61 14.68 56.74 G4A 2014 2.50 2.54 9.22 4.55 5.11 12.29 5.42 10.40 47.96 G1A 2015 1.27 1.53 6.27 4.06 3.93 11.14 9.69 11.37 50.74 G2A 2015 0.74 0.69 2.59 1.39 1.82 5.21 2.49 15.83 69.25 G3A 2015 0.29 0.58 2.83 1.32 2.36 8.07 9.59 14.60 60.38 G4A 2015 2.53 2.46 7.18 4.93 3.38 11.46 7.46 10.87 49.71 G1A 2016 0.32 1.00 5.90 3.18 5.33 10.07 7.94 11.08 55.17 G1A 2016 0.85 0.86 2.98 1.50 1.48 5.77 7.15 13.70 65.69 G2A 2016 1.45 1.36 4.19 2.90 2.18 8.50 8.01 10.90 60.51 G3A 2016 3.18 3.56 9.39 5.36 5.27 13.48 6.06 8.68 45.03 G4A 2017 0.45 1.11 6.13 4.21 10.34 7.59 8.00 9.14 53.03 G2A 2017 0.69 0.82 2.68 1.14 4.04 3.54 6.71 12.74 67.64 G3A 2017 1.82 1.56 5.21 2.90 7.69 5.92 7.54 8.24 59.12 G4A 2017 4.88 4.03 10.91 4.93 11.40 7.16 7.07 7.74 41.88 Table 12. Percentages of grades (1 to 9) by disciplines and by year 7. Conclusions This paper is destined to be the basis of a series of more insightful papers using a variety of more or less deep statistical techniques. At least, the data is entirely described. The notations are posed. A general algorithm allowing to get all relevant variables and their local restrictions (in time (year), in area (region), in gender (female, male)). First statistical trends are given, concerning incomplete data, outliers, fair- ness of representation of sub-groups in the educational data, some general evolutions of grades in the time. These statistical facts already direct to THE GAMBIAN GABECE EDUCATIONAL DATA 19 garde/year cr1-4 cr1-5 cr1-6 pass fail G1A 2012 14.94 18.67 27.43 44.03 55.97 G2A 2012 5.05 6.87 12.56 33.67 66.33 G3A 2012 6.13 9.13 16.98 29.8 70.19 G4A 2012 19.92 25.07 37.67 53.74 46.27 G1A 2013 7.26 9.71 16.43 34.64 65.35 G2A 2013 7.77 10.53 17.7 34.68 65.31 G3A 2013 5.05 6.79 15.12 33.01 66.98 G4A 2013 16.63 21.28 32.58 49.87 50.12 G1A 2014 6.2 8.54 17.51 39.66 60.35 G2A 2014 4.12 5.65 10.93 27.28 72.71 G3A 2014 9.44 11.84 19.99 43.28 56.74 G4A 2014 18.81 23.92 36.21 52.03 47.96 G1A 2015 13.13 17.06 28.2 49.26 50.74 G2A 2015 5.41 7.23 12.44 30.76 69.25 G3A 2015 5.02 7.38 15.45 39.64 60.38 G4A 2015 17.1 20.48 31.94 50.27 49.71 G1A 2016 10.4 15.73 25.8 44.82 55.17 G1A 2016 6.19 7.67 13.44 34.29 65.69 G2A 2016 9.9 12.08 20.58 39.49 60.51 G3A 2016 21.49 26.76 40.24 54.98 45.03 G4A 2017 11.9 22.24 29.83 46.97 53.03 G2A 2017 5.33 9.37 12.91 32.36 67.64 G3A 2017 11.49 19.18 25.1 40.88 59.12 G4A 2017 24.75 36.15 43.31 58.12 41.88 Table 13. Percentages of regrouped grades (credit, pass, fail) by discipline and by year interesting leads and suggest hypotheses to be validated or not. Acknowledgement The authors wish to thank the Ministry .... that al- lowed to use the data. References Crawley M. J.(2009). The R book. Wiley. James G., Witten D., Hastie T. and Tibshirani R.(2008) Introduction to th statistical learning with Applications in R. (Corrected at 8 edition 2017). Springer y yy yyy yyyy ´ ´ 20 OUSMANE SAINE , SOUMAILA DEMBELE , GANE SAMB LO , MOHAMED CHEIKH HAIDARA R Core Team (2020). R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria. https://www.R-project.org. Statistica com. https://www.statista.com/statistics/967843/total- population-of-gambia-by-gender/. Visited on 2020/11/13. Wikipedia website. https://en.wikipedia.org/wiki/The Gambia#Administrative divisions. Visited on 2020/11/11. THE GAMBIAN GABECE EDUCATIONAL DATA 21 Appendix : R codes See Crawley (2009) for an overall review on R software. #The data are i n the d i r e c t o r y gabece , which i s # o f a s u b d i r e c t o r y where i s t h i s s c r i p t . #The d i r e c t o r y should contain the f o l l o w i n g f i l e s r e l a t i o n t o #the whole population o f students , having the same lenghth #1 − mesAppurements . t x t : contains the v a r i a b l e okData with two values #”ok ” f o r complete data ” ko ” f o r incomplete data .nn #2 − the o r i g i n a l grades G1, G2, G2, and G1. #3 − the c o v a r i a t e s year . t x t , gender . t x t and r e g i o n . t x t appu <− read . t a b l e ( ” gabece/mesAppurements . t x t ” , h=TRUE) attach ( appu ) t a b l e ( okData ) Gender <− read . t a b l e ( ” gabece/gender . t x t ” , h=TRUE) attach ( Gender ) gender t a b l e ( gender ) Year <− read . t a b l e ( ” gabece/year . t x t ” , h=TRUE) attach ( Year ) t a b l e ( year ) Region <− read . t a b l e ( ” gabece/ r e g i o n . t x t ” , h=TRUE) attach ( Region ) t a b l e ( r e g i o n ) English <− read . t a b l e ( ” gabece/G1. t x t ” , h=TRUE) attach ( English ) t a b l e ( G1 ) Maths <− read . t a b l e ( ” gabece/G2. t x t ” , h=TRUE) attach ( Maths ) t a b l e ( G2 ) Sciences <− read . t a b l e ( ” gabece/G3. t x t ” , h=TRUE) attach ( Sciences ) y yy yyy yyyy ´ ´ 22 OUSMANE SAINE , SOUMAILA DEMBELE , GANE SAMB LO , MOHAMED CHEIKH HAIDARA t a b l e ( G3 ) SES <− read . t a b l e ( ” gabece/G4. t x t ” , h=TRUE) attach ( SES ) t a b l e ( G4 ) # Cleaning ============================================= # V a r i a b l e s i n c l u d i n g cases not p r e s e n t i n g missing data −1 , −2 , e t c . . genderA=gender [ okData==”ok ” ] yearA=year [ okData==”ok ” ] regionA= r e g i o n [ okData==”ok ” ] G1A=G1 [ okData==”ok ” ] G2A=G2 [ okData==”ok ” ] G3A=G3 [ okData==”ok ” ] G4A=G4 [ okData==”ok ” ] # ======================================== # y e a r l y A n a l y s i s #In the example below , you may change 2012 #by any year i n (2012 , . . . , 2 0 1 7 ) and #proceed t o the a n a l y s i s yearOfStudy=2012 regionOfStudy = ” 6 ” genderAY=genderA [ yearA==yearOfStudy ] regionAY=regionA [ yearA==yearOfStudy ] G1AY=G1A[ yearA==yearOfStudy ] G2AY=G2A[ yearA==yearOfStudy ] G3AY=G3A[ yearA==yearOfStudy ] G4AY=G4A[ yearA==yearOfStudy ] # study o f r e g i o n genderAR=genderA [ regionA==regionOfStudy ] G1AR=G1A[ regionA==regionOfStudy ] G2AR=G2A[ regionA==regionOfStudy ] G3AR=G3A[ regionA==regionOfStudy ] G4AR=G4A[ regionA==regionOfStudy ] # study o f r e g i o n and year THE GAMBIAN GABECE EDUCATIONAL DATA 23 genderAYR=genderA [ ( regionA==regionOfStudy ) & ( yearA==yearOfStudy ) ] G1AYR=G1A [ ( regionA==regionOfStudy ) & ( yearA==yearOfStudy ) ] G2AYR=G2A [ ( regionA==regionOfStudy ) & ( yearA==yearOfStudy ) ] G3AYR=G3A [ ( regionA==regionOfStudy ) & ( yearA==yearOfStudy ) ] G4AYR=G4A [ ( regionA==regionOfStudy ) & ( yearA==yearOfStudy ) ] l e n g t h (G4AYR) #proceed t o your a n a l y s i s

Journal

Statistics – arXiv (Cornell University)

Published: Dec 15, 2020

There are no references for this article.