Brenda Said:

Statistics: How do I prove a significant difference between the results of control and experimental groups?

We Answered:

You will have to use the mathematical expressions of either ratios or percentages, or both; so that the results are clearly defined in an obvious manner. The term "significant" is much too vague and undefined. Stay away from it, unless you can define it! Simply show 2 groups, 2 elections, 2 sets
of ratios and/or percentage differences in each one; and let the numbers do the talking.

Melvin Said:

Which is the correct statistical operation to use to determine significance in this situation?

We Answered:

I'm not really sure what difference calling one group the treatment and one the control would make. It doesn't really change the tests that you can do.

First of all, you kind of have to ask why you are doing a hypothesis test. We are testing whether two population means are equal, but if you have the results of the whole population, then there really isn't a need to do the test, you just compare the results. I suppose you could say something like if you were to repeat the some conditions on Election Day, then the turnout would be different each time, and so, in that sense, the turnout is random.

Now each state is going to have a different "sample" size, so I don't know what effect that would have. I guess that you are considering each state's turnout as your measurements.

I guess there are different comparisons that you can do here. You can compare turnouts from between the "experiment" and "control" groups, and you can compare pre and post 1993 results.

I guess since that's the case, then you can try to use a two-sample t-test on the comparison between the "experiment" and "control" groups. That is because these would be independent samples. However since one groups has four "observations", then you would have to be sure that the population of turnouts is normally distributed.

I don't have any experience with voter turnouts, so I can't really say either way. If not, you may want to look into the rank sum test. That is a nonparametric test, so you don't need the normal population assumption, but you do need to assume that the populations have the same shape of distributions (so both are symmetric or both are skewed to the right for example).

Now for the pre and post 1993, these would be paired by the state that it's coming from. So you could possibly do a paired t-test, but the same normal caveat applies here as well. The nonparametric equivalent is the signed rank test.

I'm not going to put my reputation on the line on this (not that I really have one). I would suggest speaking to a statistician on this one. Maybe try to seek one out at your university/college.

Hugh Said:

Questions about U.S. voter registration!?

We Answered:

Reason Enforcer is right. You can register anywhere but you're supposed to be mature and intelligent enough to pay attention and know what address you're using.

But if you want to fix this now, decide where you want to be registered - at home or at school.

Then complete a new voter registration form with that address and whatever party you want, or none, or whatever.

Go to www.usa.gov.
Click "voting and elections"
Click "register to vote by mail

Print the form that appears.
Complete the form, make a photocopy for your records.

Follow the instructions to mail the form to the correct address for your state.