How to measure the impact of a subject on student performance?

#1
Hi,

I need to find the impact of mathematics on student performance. I have only the student exam marks for mathematics and all the other subjects. What are the statistical methods that can be used for this situation other than correlation and regression? According to my point of view, I think it would be better to use multivariate analysis.

Thanks
 
#2
I would need a bit more information about your dataset to give you an idea of where you are heading.

In particular, examining students testing data is where the large portion of my research interests are concerned. I will caution you that there are quite a few issues that you should be aware of.

Questions-
1. is the data longitudinal or cross sectional?
2. are there missing data values?
3. are they all taking the exact same test version?
4. what are the characteristics of the test score distribution? (are there floors and ceilings?)

I at least need that sort of information before analyzing the exam.
 
#3
Hi,

The study was conducted for two batches from 7 different disciplines at the Faculty of Engineering in a University. Student exam marks of mathematics for all 4 years is available. All students have common mathematics courses only in first year and in other 3 years some disciplines may have different math courses. Student exam marks of other courses (courses are vary according to their discipline) for final 3 years are also available (first year is not).

Missing values are exist due to the absence of a student in exam (may be for 1 course or more than 1). two batches get different test as they are in two academic years. Exam marks range is 0 to 100.

Since courses are different in discipline wise, I hope to do the analysis separately for each discipline. That is, the impact of mathematics on student performance is to be find with respect to their discipline.

What are the statistical methods that can be used for this situation other than correlation and regression? According to my point of view, I think it would be better to use multivariate analysis like canonical correlation.

Please give your suggestions.

Thanks
 
#4
The first thing you need to do is to standardized those test scores. That would probably be my first step. In this case, what you are comparing is a subjects mathematics performance (please also note that there is a difference between performance and ability) relative to his or her peers. This would allow you to compare different tests across different years and different courses.

Once you have that established, you can use a longitudinal model to examine how a students relative performance on mathematics tests corresponds to his/her overall performance.

One other question, how are you defining over all performance? Do you mean in terms of performance in other classes, a GRE score, ranking among peers? What is your metric to define performance? That will determine how you will structure your analysis.