This project is formed from this Kaggle page and conducts hypothesis testing to observe the effects of days of the week, year, teachers, class start time, and subject material upon GPAs.
If the class meets on weekdays, we did not find an effect upon GPA. However, we found classes that meet on weekends to be statistically distinct at a 95% confidence interval. After factoring in the Bonferroni correction, we found that not only are weekends still distinct from the weekdays, but Fridays are also distinct from the other weekdays. Day of Week and GPA
We did not find a significant difference in GPAs between these two school years, which were a decade apart. Grade Inflation from 2006/2007 to 2016/2017
We found that earlier and later years in the data set that were statistically distinct from the overall data set at a 95% confidence interval. Even factoring in the Bonferroni correction revealed some differences. Each Academic Year Compared to the Others
We did originall find some teachers (~12) in the math department that graded differently from the rest of the department. We did not find that the others graded differently on a statistically significant basis. However, after factoring in the Bonferroni correction, we found that only 3 teachers graded differently from the rest of the department. Math Teacher Grading Consistency
We did not find a difference in GPAs for classes that start in the morning vs classes that start in the afternoon. Time of Day and GPA
We found that STEM classes have a different GPA than non-STEM classes to a statistically significant degree at a 95% confidence interval. STEM vs Non-STEM GPA
In order to access this data, please make a Kaggle account and download the zip file here.
This project assumes that the user will fork and clone this project locally and that the data will be downloaded into a data subfolder in the project directory.
The original presentation and results of this project did not check for multiple tests and was written in a way that only works on Macs.
Currently, the individual result summaries do factor in the Bonferroni correction where appropriate, but the technical setup instructions still could use some work to walk through setting up and running on PC.
If you are curious about technical documentation for conducting these hypothesis tests, please be aware that you will:
If you want to see the individual summaries, you can open a Jupyter Notebook server in the base project directory and navigate to each individual hypothesis folder to read the notebook files.