# Thread: What statistical method should I use !

1. ## What statistical method should I use !

I would appreciate if any one could give my some advice of what statistical method/model I should use when analysis the following data

Pre-test Post-Test

Intervention Group

Control Group

From my limited understanding (very limited!) the Intervention group is a repeat measures so I need some statistical method to identify if there is a significant difference between pre & post. I also need identify if there is a significant difference between the intervention and control which are independant groups.

Any help will be greatly appreciated !

2. ## Re: What statistical method should I use !

The pre post variable is your repeat. The intervention control variable is your between groups variable. So, from what you have written, you have two independent variables and this is sometimes referred to as a mixed design because one repeats and the other does not.

Assuming you are only measuring one dependent variable, you have a 2 x 2 mixed design. You would use an ANOVA and you would be testing 1) the main effect of time (pre/post), 2) the main effect of group (intervention/control), AND 3) the interaction of time*group. In this design you typically only care about the interaction. Most researchers want to see that the intervention group has had a better time effect as compared to the control.

3. ## Re: What statistical method should I use !

However in reading your response it has stimulated a few other concerns I have with the appropriate statistical method.

Firstly as previously stated there is a control group and intervention group (independent) and each group will be tested pre and post (dependent).

They will be tested pre and post however on a number of different measures. It was my intent as to address these individually.

However is it possible (or should I) use some type of method the address all these variables simultaneously a e.g. 2 x 15 if there is such a thing.

Then to throw an even bigger spanner in the works there is a phase 2. where the original intervention group gets split into a control group and intervention group.

Again any help is appreciated.

Cheers

4. ## Re: What statistical method should I use !

MANOVA can handle the same 2 x 2 mixed design. It will allow you to combine more than one dependnet variable. If for example, you have measured 15 different physiological variables on humans, you would run one MANOVA including the cholesterol variables, one MANOVA including the blood pressure variables, one MANOVA including the body size variables, etc. And if you had a single variable that did not really fit into a group of variables, you would just run a univariate ANOVA on that one.