multiple regression using sas

21
Predicting Academic Performance of an Elementary School Submitted By Ankush Roy Ashitha VS Koushik Rakshit Krishna B Roma Agrawal 1 (Multiple Regression using SAS)

Upload: krishna-bollojula

Post on 14-Apr-2017

371 views

Category:

Data & Analytics


1 download

TRANSCRIPT

Page 1: Multiple Regression using SAS

Predicting Academic Performance of an Elementary School

Submitted ByAnkush RoyAshitha VS

Koushik RakshitKrishna B

Roma Agrawal1

(Multiple Regression using SAS)

Page 2: Multiple Regression using SAS

Data Description

2

• Dataset contains performance data of 400 elementary schools from the California Department of Education.

• Attributes like average class size, parent’s education, students opting for free meal etc.

• Many attributes have missing values, outliers which needs to be fixed before using this in analysis.

Page 3: Multiple Regression using SAS

3

Objective

• To find important attributes that majorly affect the academic performance

• Using this attribute, build a model that can predict the overall academic performance of school

• Verify the model before accepting that.

• Create an automated excel sheet using above regression model, which will give the performance taking attributes as an input.

Page 4: Multiple Regression using SAS

4

Finding Important Attributes, Building Model and its

Verification

Page 5: Multiple Regression using SAS

5

Attributes and their ImpactAfter statistical analysis, we found below 5 attributes to be most important and on which performance depends.

api00=459.713-2.899(ell)-3.108(mobility)+3.688(acs_46)+3.377(grad_sch)+2.332(full)

Regression Equation

Page 6: Multiple Regression using SAS

6

Model Verification

MAPE (Mean Absolute Percentage Error)MAPE=Mean(ABS((Actual value – Predicted

value)/actual value) )*100(It should be <10)

For our model, MAPE=8.9140897

Adjusted R-SquareFor our model, Adj R-Sqr=76.5%

Page 7: Multiple Regression using SAS

7

Residuals are normality distributed (as per graph and also with Anderson-Darling test)

Page 8: Multiple Regression using SAS

8

Attribute-wise Interpretation and Recommendation

Page 9: Multiple Regression using SAS

9

English Language Learner• ELLs have negative impact on overall performance of

school means decreasing this will improve performance.• ELLs more means students are coming from different

geographies where English speaking is less and they are poor as they haven’t learnt English speaking.

• There will be a challenge in communicating to these students.

• Schools can work on this, to improve its performance by providing special coaching to ELLs to ensure students master in English.

• Recruit teachers who can better understand the need of these ELLs.

Page 10: Multiple Regression using SAS

10

Percentage first year in school(Mobility)

• %age of students who are not completing their studies in school.

• It is negatively affecting the performance, as students are studying for some time and then they move. Decreasing mobility will improve the performance.

• It affects students also, as they will not be able to concentrate on studies.

• One of the reason for mobility can be residence change of family.

• School should consult with parents and make them understand to keep their child in same school.

Page 11: Multiple Regression using SAS

11

Average Class Size 4-6• It is positively associated with performance of

school means increase in this will increase the overall performance also.

• It can be improved by – decreasing mobility– Increasing promotion of students from one grade

to another grade– Increase in students taking admission each year.• This is the direct indicator of overall performance

of a school. But without controlling other variables, only increase in this will not help school in increasing performance.

Page 12: Multiple Regression using SAS

12

Parent’s Grad School• It has positive impact on increasing performance of a

school.• If parents will be educated they can take care of

children’s studies at home as well. They will take interest in their homework, they will ask for marks etc. which will help child a lot.

• Children can learn many things from their parents apart from school.

• School can start some initiative for parent’s involvement like monthly parent’s-teacher’s meeting in which teachers can discuss with parents about their child’s studies.

Page 13: Multiple Regression using SAS

13

Percentage Full Credential• This is percentage of full time teachers in school.• It has positive impact on performance of school as

they are the source of knowledge for the students. In case of absence of temporary teachers, students would be sitting idle.

• School should recruit more full time teachers and provide them good working environment so that they can give their 100% in sharing their knowledge with students and help them in learning better.

• Shortage of teacher can also be a reason for student’s mobility and low performance.

Page 14: Multiple Regression using SAS

14

Dashboard Creation

Page 15: Multiple Regression using SAS

15

api00=459.713-2.899(ell)-3.108(mobility)+3.688(acs_46)+3.377(grad_sch)+2.332(full)

Regression Equation

**By entering values for 5 attributes, api00 (performance) can be calculated

Page 16: Multiple Regression using SAS

16

Increasing ELL, decreases the performance value(API00)

Impact of English Language Learner

Page 17: Multiple Regression using SAS

17

Increasing Mobility, decreases the performance value (API00)

Impact of Percentage first year in school(Mobility)

Page 18: Multiple Regression using SAS

18

Increasing acs_46, increases the performance value (API00)

**But this should be increased by controlling other factors. Simply increasing class size will not help in improving performance.

Impact of Average Class Size 4-6

Page 19: Multiple Regression using SAS

19

Increasing grad_sch, increases the performance value (API00)

Impact of Parent’s Grad School

Page 20: Multiple Regression using SAS

20

Increasing full, increases the performance value(API00)

**This is the attribute which is totally in control of school. So increasing this will help a school in increasing performance.

Impact of Percentage Full Credential

Page 21: Multiple Regression using SAS

21

Thank you…!!!