Skip to content

Root Cause analysis to check for disparity in testing performance across different jurisdictions across US

Notifications You must be signed in to change notification settings

theheisenberg10/Root-Cause-Analysis-for-Testing-Service-Provider-

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

7 Commits
 
 
 
 

Repository files navigation

Root-Cause-Analysis-for-Testing-Service-Provider

Visualized variation in test scores among different jurisdictions in the United States explaining 8 KPIs in Tableau

Identified causal effect of internal study resources on student test performance using propensity score matching

Improved overall average candidate performance by 10% delivering recommendations for improving exam preparation strategies in underperforming regions

Technical Approach

We did two analyses in this project. Firstly, we got the top and bottom 5 jurisdictions according to the average testing scores. Then we ran a regression model on these jurisdictions only and found out factors that drive the score up or down like number of practice tests, use of different study materials, etc.

Results

We found study resources that were used by high scoring jurisdictions and if also used by the bottom ones can significantly improve their scores. We also did a dropout analysis on the data. We found out that many students were passing in 4 subjects but failed to earn a credential as they could not pass the 5th subject which was math in majority of the cases. This was bringing the credential rate of the company down.

Testing

We used causal inference techniques (T-test) to see if there was a significant variation in scores between jurisdictions and then adjusted our analyses accordingly. We identified intriguing trends and clustered them into groups using candidate level information such as a student's scores, demographics, and internal study resource utilization statistics. Then we regressed the candidates' scores on the dependent variables to see what helped them score higher and what pulled their score down.

Conclusion

We identified intriguing trends and clustered them into groups using candidate level information such as a student's scores, demographics, and internal study resource utilization statistics. Then we regressed the candidates' scores on the dependent variables to see what helped them score higher and what pulled their score down. We discovered that in certain places, employing a specific sort of study material (audio resources) was beneficial, whilst in others, a different strategy (personal tutoring) was beneficial

About

Root Cause analysis to check for disparity in testing performance across different jurisdictions across US

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published