Help with some statistics

2000AD · Post by **2000AD** » 2006-05-04 11:49am

I've got 2 variables and 5 different results.

To see if the variables are predictors i did a Regression analysis for each fo the 5 results with the 2 variables as predictors. Was that the right thing to do?

What if the results are not independant?

fnord · Post by **fnord** » 2006-05-05 03:20am

A bit more information could be handy. I'll call the variables x and y, for want of better labels.

Before you did your regression, did you do any scatterplots, and if so, what relationship did they suggest? Linear? Quadratic? Exponential?

What were you expecting your results to be independent from/of?

Kuroneko · Post by **Kuroneko** » 2006-05-05 03:37am

The right thing to do to accomplish what, exactly? You don't state your goal. Five points isn't much; anything beyond a quadratic is all too likely to generate false dependence (one can always get an exact fit on a fifth-degree polynomial). About the simplest test of independence you can do is to calculate the covariance; Cov(X,Y) = 0 is necessary, although not sufficient, for independence.

2000AD · Post by **2000AD** » 2006-05-05 08:31am

Sorry about the lack of infomation, but i was in a rush. It doesn't matter now as i've handed the work in.

Next time i need stats help i'll remember to provide more info.