1
$\begingroup$

I have about 40 variables for each subject in a human population. For each time period, people join and exit the study. As a made up example, I want to see whether there are increases in average spending on movies as time progresses. The problem is that my population is very volatile, there could be 15% male in one time period and 99% male in another. Given that this is the case, how can I figure you whether the increase I observe is due to actual increase, population change or just variance?

What I'm looking for is what subject I should be learning to address this problem. A particular textbook on clustering, regression? Or something like that. I cannot change or resample the data I'm given and I'm looking for something that's masters or bachelors level.

$\endgroup$

2 Answers 2

1
$\begingroup$

Look at bootstrapping first.

For example, you would do 100 random permutations of your data set, and check how many false findings you get with your approach.

$\endgroup$
1
$\begingroup$

me thinks this is a pretty good case for a decision tree analysis. So i assume your 40 variables dont include the time period but that you have decent markers for your time period (P1 , P2 etc :) ) and you could use average money spent as a decision label. What the tree is going to do is tell you , for instance, during P1 if the subject was male and if he had a car and if he had a wife his average spending was > 200 bucks ..the tree will do this for all periods and then its a simple matter of visual interpretation for you to figure out the anwer to your problem..hope this helps

$\endgroup$

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.