2
$\begingroup$

I have a simple set of data that looks like

key     thingID     countOfThing   NumberBadThingsHappen
1       xxx         64             200
2       xxx         20             10

What are some applicable tests to see if countOfThing and NumberBadThingsHappen share any type of relationship? The first two columns serve to uniquely identify the last two. I don't anticipate this being a panacea for any real analysis, just looking for a place to start. Feel free to edit I've tagged this incorrectly.

$\endgroup$
  • 2
    $\begingroup$ I would suggest changing your strategy from testing to exploration. Start by summarizing and visualizing your data. A natural beginning to visualization (for the purposes of identifying relationships) is a scatterplot. (When you're still at the stage of wondering what type of relationship two variables might have, formal statistical tests give misleading results. Search the literature for "data snooping" and "data dredging".) $\endgroup$ – whuber Mar 27 '13 at 13:13
  • $\begingroup$ One of the bad things about learning something on your own is, sometimes you don't know what things are called! Thanks :) $\endgroup$ – wootscootinboogie Mar 27 '13 at 13:16
2
$\begingroup$

Note there are different types of correlation (Pearson and Spearman), and just because Pearson $\rho=0$ doesn't mean there is no relationship (there are nice examples on wikipedia for this). Hence, as suggested in the comments, you should make a scatterplot using the two variables.

Then you can compute the correlation coefficient that seems most appropriate.

Another quick way to test for dependence is to run a linear regression (OLS) of one variable on the other. The t test of the slope coefficient tests for a significant relationship. The sign of the slope parameter will be identical to that of the correlation coefficient.

$\endgroup$
0
$\begingroup$

Is there some reason not to use Pearson's correlation for quick analyse? And do some simple eplonatory analysis with data visualization (scatterplots etc).

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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