Hello there,

I'm a STATISTICA user and I have a question about the p value the program calculates in ANOVAs. As the majority of statistical programs, among other index such as F, MSE, df, etc, STATISTICA computes a p value for main effects and interactions.
I have been told that, similarly to the t-test, in case the computation is based on a two-tailed hypothesis I am allowed to split up the given p value.

For example, let's say STATISTICA finds a main effect with a p = 0.0984.
If the variable fits onto a two tailed hypothesis I can consider the result significant, because my p value (divided by two), is now p = 0.047 (< .05).

Thus, the question is: am I allowed to use this procedure?
Do you know whether STATISTICA returns a two-tailed or a one-tailed p value?
Thank you very much in advance.
Best
Michele

Check the null hypothesis of ANOVA. and t test.
I hope you know about the null hypothesis one tailed t test( left or right)?. So what could be corresponding change in ANOVA?

Hi Vinux, thanks for your replay.
However, I'm not sure I've got it right.

Check the null hypothesis of ANOVA. and t test.
I hope you know about the null hypothesis one tailed t test( left or right)?

I know what a null hypothesis is (one tailed or two tailed)
For example (one-tailed hypotheis) I expect that the condition "A" received higher ratings than condition "B" (H1). My null hypothesis (H0) is that there is no difference between the two conditions (or the condition "B" receives higher ratings)

So what could be corresponding change in ANOVA?

:confused: Sorry but I do not understnad what you mean...

Cheers
M.

Ok..
What will happens in ANOVA.. there will be A ,B &C
H0: mean(A)= Mean(B) = Mean(C)..

And in ANOVA we use F( it will be always +ve). when we do ANOVA for two groups , the F will be same as the square of T statistics.. So here we don't consider the sign.
Here the p-value(ANOVA)is two tailed.


For example, let's say STATISTICA finds a main effect with a p = 0.0984.
If the variable fits onto a two tailed hypothesis I can consider the result significant, because my p value (divided by two), is now p = 0.047 (< .05).

Thus, the question is: am I allowed to use this procedure?


Answer is No.

EDIT: But in 2-factorial experiment.. we can consider the half of the p-value( if we are interested verify the one sided effect). This will be justified by considering testing the main effect using t test.. But most of the cases ppl go by the two tailed pvalue.

Thank you vinux. :tup:
Now it makes sense.

Cheers
M.