Did a quick search and none of the stats threads dealt with this specifically...also I haven't been in school for years and only took a couple statistics classes so sorry if this is a dumb question.

Anyways, say I have 2 sets of numbers (around 65 in each set) that are both prices for something over a 3 month period, but from different sources. The numbers are all fairly similar (between the 2 sets, they change drastically over time), but never exactly the same.

What I'd like to do is be able to say something like, "Although the values are clearly not exact, their differences are not statistically significant" (assuming that's the answer I get).

It looks like a PValue might be what I'm looking for, which I can't figure out how to calculate...I have been able to get the TValue and it passes the T Test, but since these are prices over time I'm not sure a TTest is exactly what I'm looking for

4/30/2010 11:04:16 AM

Quote :

"that are both prices for something over a 3 month period"

Sounds to me like you are on the right course with the T test. I'd think a T test for paired differences if its the same subjects across time.

4/30/2010 11:40:08 AM

nm

[Edited on April 30, 2010 at 11:49 AM. Reason : nm]

4/30/2010 11:49:28 AM

I would do a two-tailed t-test, wherein your null hypothesis is average1 = average2.

4/30/2010 11:49:57 PM

A couple questions:

- Would you expect that both sets have equal variances?
- Are you interested in testing whether the means of these two data sets are not equal?

If you want to test a difference in means, a t-test is what you should use.

H0: Mean1 - Mean2 = 0 vs. Ha: Mean1 - Mean2 =/= 0

If you have equal variances, use the standard student's t-test:

http://en.wikipedia.org/wiki/Student%27s_t-test#Independent_two-sample_t-test

If you suspect inequality of variance, use Welch's t-test:

http://en.wikipedia.org/wiki/Student%27s_t-test#Independent_two-sample_t-test

Look up the value in a t-table to get a corresponding p-value, and reject the null accordingly.

5/3/2010 9:59:59 AM