This analysis is part of the final project in the course STAT 6021: Linear Models for Data Science.
The project is Regression Analysis on Graduate Admissions Data in R and I performed this as part of a group of 5 students. A brief background on the project is given below.
Background
Each year, over 3 million college applications are filed in the US by about 750,000 students, an average of 4 applications per student. Each of them comes with a certain element of randomness or chance. The intended meritocracy inherent in college admissions gives way to uncertainty, doubt, and anxiety, even for students with exceptional credentials. Not all colleges are transparent about their admission processes and so it becomes tough for a student applicant to gauge whether he or she can get admission into an institution. In this project, we demonstrate how regression analysis can be used on a sample dataset to ease this process. Universities can use similar analysis on their data to help with the admissions process, and students can use this analysis to determine how likely they are to get an admit given the strength of their profile.
There are many factors that influence admission decisions. And while colleges rely on more than quantitative data to make admissions decisions, quantitative data can show us in a concrete way many things that qualitative data cannot. Even though it's tough to understand and estimate how these factors are truly judged and filtered by colleges, we do know that some of the factors such as CGPA and GRE scores can weigh heavily on determining acceptance. Metrics such as these scores can be leveraged in form of data and be analyzed to gain insight into admission trends and can help students in shortlisting universities with their profiles saving time, effort and money that goes into the exhaustive application process. The predicted output can also give them a fair idea about their chances for admission to a particular university. The scope of such analysis can also be extended to help college institutions answer questions such as – "Do we know that standardized tests are a valid predictor of success in admission at our institution?"
We have adopted a data-driven approach towards quantifying the probability of successful admission or enrollment into college institutions dependent solely upon certain quantitative factors. Our objective with this analysis is two-fold. The first objective is to understand what factors are significant and relevant in determining enrollment and to what degree. In doing so, we draw inferences about relationships between the factors and identify any dependency that exists between them. The second objective is to build upon the first to develop a predictive model based on regression that can compute probability of an admit.
| Visit: | Project link |
| Project Report: | See |