# Hypothesis Test Problem

• Apr 6th 2013, 05:10 PM
jeff2222
Hypothesis Test Problem
Hello,

I have a problem involving making a Hypothesis Test for a treatment being performed. The outcome is measured with a health rating FIM. There are 203 patients before treatment and 203 after treatment , each has a health measurement FIM. I found a sample mean of 41.8 before treatment with a variance of 58.5, and a sample mean of 61.5 after treatment with a variance of 99.7. I want to perform hypothesis testing on this. I made the null hypothesis that the average after the treatment was the same as the sample mean before the treatment (41.8). Was that wrong to do?

When finding the z-score of the result I got approx. 28, which I do not think is correct because of how high it is. From what I understand I have too many samples for a two sample t-test correct? But nonetheless when I perform that I get a score of approx. 22. Which I think may also be incorrect.

Thanks,

Jeff
• Apr 6th 2013, 05:16 PM
chiro
Re: Hypothesis Test Problem
Hey jeff2222.

You don't have too many samples for a t-test. Using the normal approximation is valid since you have a lot of samples.

Can you show us how you calculated your test statistic? Are you using a paired t-test? (Are the relationships between each pair strong enough to consider a paired test?)
• Apr 7th 2013, 10:59 AM
jeff2222
Re: Hypothesis Test Problem
Hey thanks for the response,

(standard deviation of FIM at discharge)/sqrt(total number of patients) = 0.7006

(mean of FIM at discharge - mean of FIM at admission) = 19.6108

19.6018/0.7006 = 28

For the t-test I used:
t = (mean of FIM at admission - mean of FIM at discharge)/ sqrt((SD1^2/203 + SD2^2/203))

(41.8 - 61.5)/ 0.883 = -22.2

The patients at discharge are the same patients at admission, so there should be a strong relationship.
I think I may understand what the problem is, I need to be calculating the mean differences between the tests. Is that correct?
• Apr 8th 2013, 03:37 AM
chiro
Re: Hypothesis Test Problem
Are you using a paired t-test? You should probably switch to one if you are not already doing so.
• Apr 8th 2013, 02:55 PM
jeff2222
Re: Hypothesis Test Problem
I wasn't. I now am. Thanks for your help.

For anyone else reading this, the method for the paired t-test is explained well here: https://www.youtube.com/watch?v=vB1OmEY5Rcw
And the matlab function ttest has options to be used for a paired t-test.