correlation between 2 groups

[safig]

Hi

I have given questionnaires to 50 participants, and i want to see if there are any differences in the mean values between those participants with an even compared to an odd questionnaire number. So i have split the participants into 2 groups- group A(even nos) and group B(odd nos). I have the means for group A and group B.

I was wondering if there is a correlation caclulation i can do on these two means, to see how different the means are from each other?

Many thanks

[WeeG]

I wouldn't use correlation calculation, it ain't appropriate, if you want to compare 2 means, you should think more in the way of the t-test, or the equivalent non parametric text, in case the assumptions of the t-test will not be valid.

[safig]

HI

Many thanks for your reply. But i didn't quite understand- when would the assumptions of the t test not be valid? And would you be able to suggest a non-parametric test or a website where i can find out more.

Many thanks

[Dragan]

Hi

I have given questionnaires to 50 participants, and i want to see if there are any differences in the mean values between those participants with an even compared to an odd questionnaire number. So i have split the participants into 2 groups- group A(even nos) and group B(odd nos). I have the means for group A and group B.

I was wondering if there is a correlation caclulation i can do on these two means, to see how different the means are from each other?

Many thanks

Well, yes, there is (sort of).

First, you take your data and put it into a single column called Score.

Second, you create a second column called Group. You assign 1’s to the scores from the first group and 2’s to the scores from the second group.

Third you compute the (Pearson) correlation (r) between Score and Group.

Fourth, compute t as:

t = (r*Sqrt[N-2]) / Sqrt[1 – r^2]

Compare your computed t to the critical value of t on N – 2 degrees of freedom.

Note: This process is logically equivalent to conducting a two-independent samples t-test.

[safig]

Hi

thank you for your reply. I was looking at the pearson calculation on excel help- does one set have to be independent values and the other dependent?

Also, is N the number of values? I did the calculation and it returned t = 0.141756158, how do i compare this- didn't quite understand the N-2 degrees of freedom


many thanks once again

[WeeG]

before using the t-test, you must verify that the scores of both groups have the normal distribution ( or at least approximately normal ), and that their variances are equal. if this is not the case, I think it would be better to use the Wilcoxon-Mann Whitney test.

[safig]

Hi

Thanks for your reply. How do i confirm they have a normal distribution? - is it just a case of plotting the data on a graph and looking to see if the distribution forms a bell-shaped curve?

I've checked the variance for group A and group B. Group A is 4.3 and Group B is 4.01. So i guess this rules out using the t test?

I've done a t-Test in Microsoft Excel 2007: Two-Sample Assuming Unequal Variances

Variable 1 Variable 2
Mean 3.295250896 3.337365591
Variance 4.247571555 3.960977676
Observations 93 93
Hypothesized Mean Difference 0
df 184
t Stat -0.141756158
P(T<=t) one-tail 0.443713852
t Critical one-tail 1.653177088
P(T<=t) two-tail 0.887427704
t Critical two-tail 1.9729405

I would be grateful if you could help me interpret what this means? Is it the right test?

Many thanks

[gianmarco]

Hi,
just few hints:

1) attached article on Welch t (t-test for unequal variance)

2) my Excel template for comparing two groups (see my previous post http://talkstats.com/showthread.php?p=25323#post25323)

Hope this help,
Regards

Gm

[safig]

Hi

thanks for your reply. I've managed to do the unequal variance t test (which i understand is the same as the Welch test). Problem is, i'm not sure how to interpret the results. Would you be able to help at all? Results are below

t-Test: Two-Sample Assuming Unequal Variances

Variable 1 Variable 2
Mean 3.295250896 3.337365591
Variance 4.247571555 3.960977676
Observations 93 93
Hypothesized Mean Difference 0
df 184
t Stat -0.141756158
P(T<=t) one-tail 0.443713852
t Critical one-tail 1.653177088
P(T<=t) two-tail 0.887427704
t Critical two-tail 1.9729405



Many thanks

[gianmarco]

Hi,
would you mind to let me inspect your data?

You can attach them as a reply or as a private message to me.

May be I can give you a better comment.

Regards,
Gm

[safig]

Hi

I'm afraid i'm not allowed to send the data, sorry. Is there any chance you would be able to comment on the results as they are- or tell me what sort of values i should be looking at?

thanks

[Dragan]

Hi

thanks for your reply. I've managed to do the unequal variance t test (which i understand is the same as the Welch test). Problem is, i'm not sure how to interpret the results. Would you be able to help at all? Results are below

t-Test: Two-Sample Assuming Unequal Variances

Variable 1 Variable 2
Mean 3.295250896 3.337365591
Variance 4.247571555 3.960977676
Observations 93 93
Hypothesized Mean Difference 0
df 184
t Stat -0.141756158
P(T<=t) one-tail 0.443713852
t Critical one-tail 1.653177088
P(T<=t) two-tail 0.887427704
t Critical two-tail 1.9729405



Many thanks

Safig: Quite frankly, by the looking at your observed Very Small t statistic (t=-0.141756), I really don't think that any alternative procedure is going to change the conclusion that you're means are not statistically different.

[safig]

Hi

Great, so the means from the two groups are not statistically different- which is what i expected. What sort of t value would you need for the means to be statistically different.

Many thanks for ur help everyone

[gianmarco]

Hi,

Please find attached a PDF file with the output of the t-test performed with Sigmaplot.

The program is interesting since it explain the results.

The test is performed twice: one for two groups with significant mean difference; one for groups similar in mean value.


Hope this help,
regards
Gm

[safig]

Hi

thanks for your reply. I will try and download sigmaplot, and hopefully that will help me to make sense of the results!

thanks