# How can I determine if two things are statistically correlated?

Asked by bgauts (9) April 8th, 2012

I am writing a research paper and trying to determine if the way that a Congressional district voted for president in 2008 (for Obama or McCain) is correlated to how the Democratic incumbent did in their 2010 Congressional race. Those are the two variables.

My hypothesis is that districts that voted for McCain in 2008 did not re-elect their Democratic member of Congress in 2010.

How do I test this statistically? My professor said I could do a Chi square but I’m not sure how to go about doing that.

Observing members: 0 Composing members: 0

Welcome to Fluther.

Your hypothesis can be easily checked by direct observation. That is, list the districts that voted for McCain, then filter out those which already had a Republican Representative. You’ll have a list of all districts who voted for McCain and had previously had a Democratic Representative.

All you need to do then is re-examine how those districts voted in the Congressional election and you’ll know whether your hypothesis is correct or not.

First of all, I wonder how large your primary data set will be. How many districts voted for John McCain and had previously had non-Democratic Congressional representation?

EDIT: First of all, I wonder how large your primary data set will be. How many districts voted for John McCain and had previously had Democratic Congressional representation?

It doesn’t seem (to me) to be a particularly valid study of anything. It’s like suggesting a “hypothesis” that people who eat at a particular restaurant normally order edamame as an appetizer when they order the crunchy tuna rolls as an entrĂ©e. Well, either they do or they don’t. You count the orders of each and see if it’s true or not. Not much of a study, I think.

CWOTUS (24753)

How many districts voted for John McCain and had previously had non-Democratic Congressional representation?

@bgauts should have an answer soon.

jaytkay (25763)

@CWOTUS @jaytkay well I’m looking at districts who voted for Democratic Congressional representation in 2008 despite voting for McCain in the presidential election. I’m comparing those district to ones who voted for Democratic Representation in 2008 and also voted for Obama.

I’m comparing these two types of districts in the area of 2010 Congressional elections to see if they reelected their Democratic representative or if they went Republican. If the district voted McCain in 2008, they should be more likely to change to a Republican Congressman in 2010. But I want to show this statistically, such as with a Chi square, but I have no idea how to go about doing that?

bgauts (9)

Your professor is correct. In order to do a correlation (linear regression) analysis, you need the two variables to increase linearly. The data you have won’t fit into a correlation.

Because of this, the best way to evaluate them statistically is using a chi-square test. You can find a chi-square calculator here.

If you are not sure what a chi-square measures or what the outcome means I can help you with that too.

nikipedia (27439)

Doesn’t Excel calculate Chi-Squared?

6rant6 (13667)

or