0
$\begingroup$

I want to predict the performance of several students by using their grades during the first three years of high school in order to produce the grades of the next three years. I'm interested in any suggestion of regression methods.

Here's an example of my data

         Year1 Year2 Year3 Year4 Year5 Year6
Student1 58    14    58    12    57    64
Student1 74    24    99    13    11    19
Student1 63    25    35    45    78    77
Student2 13    15    89    99    98    100
Student2 78    84    80    85    100   100
Student2 12    45    48    46    30    44
Improve this question
$\endgroup$
1
  • $\begingroup$ Is it important to predict each year individually or would you be satisfied knowing that the mean of years 1-3 predicts the mean of years 4-6? $\endgroup$
    – David Lane
    Commented Mar 22, 2017 at 17:07

1 Answer 1

Reset to default
0
$\begingroup$

As the predictors are all auto-correlated i.e. what a student achieves in Year 2 will be correlated with what they achieved in Year 1 you'll need to look at some form of time series analysis such as ARIMA modelling. I'd suggest taking the average score achieved by all students for each time period and fitting an ARIMA model. You can then forecast future grades for each student by imputing their particular time series of grade averages.

If you use R then this is a good resource to get started:https://www.otexts.org/fpp/8

Improve this answer
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.