Might be a silly question about t-tests but...

[Zzzz]

...i really need some advice about how to go about comparing some means. Basically, I have the means and standard deviations for three different groups (sample a, b and c) on one particular variable (a mood measure). I want to compare whether there are significant differences between each sample (different groups of people's scores on this one mood measure). However I can't work out how to input just the means and SD's into a t-test on SPSS As far as i can make out it seems i need the individual data, but i don't have this, its just the mean and SD's. Is there a way of doing this manually perhaps?
Does anyone have any advice about how I can do this?

Thanks
Zz

 

[JohnM]

If you have the means, standard deviations, and sample sizes, you can do t-tests without the individual data points. Do you have the sample sizes?

 

[Zzzz]

Hi John, yes I do have the sample sizes. The n sizes are 47, 1792, 25! So there is quite a difference between. How do i do the t-test? I'll need step by step instructions!!! Much appreciate your help

 

[JohnM]

Here's a link to an online tool:

http://home.clara.net/sisa/t-test.htm

 

[Zzzz]

Great thanks John. Although, when I put my data in I get a p value of '0' (not 0.00 with any other values). What would this mean?

 

[JohnM]

It means that the difference between the means, relative to their standard deviations, is very large. Therefore, the likelihood of them coming from the same population is extremely small (the p-value is the probability of getting the observed t-statistic if in fact Ho is true).

 

[Zzzz]

O right ok, I thought it might mean that but I wasn't sure. I thought that given the large differences in the sample numbers it might have shown a '0' because of a massive error, but perhaps it just means as you said the difference is very large. Which is great for what i'm looking at! Always good to get a significant result.

Next thing to quiz you on... I'm having to also do mediational analyses. I have an IV, DV and mediator. The regressions are all significant (3-ways). When the mediator is included with the IV as predictors of the DV, the mediator is not significant, but then neither is the IV and DV relationship. Therefore it looks like whilst the mediator shows a non significant effect, it still affects the IV and DV relationship as it knocks out the previously held relationship. What does this mean? does it still mean that the mediator has an effect given that when its present it renders the IV-DV realtionship not sig'.
Once again, your help and advice would be much appreciated.
anna

 

[JohnM]

It's a problem called multicollinearity. When you have two IVs that are significantly related to a DV, but are also significantly related to each other, they "rob" each other of significance.

The model you've developed isn't necessarily wrong, it's just not able to distinguish between the subtle differences in the contributions of each IV in predicting the DV.

The solution is to just explain the theory and results in terms of the individual correlations, and not a model with 2 IVs.

 

[Zzzz]

Cool! Once again big thanku John!! I'll probably have more questions...
...and thanks for the big word multicolliner....etc!

Anna