Help! What statistical analysis test should I use?

#1
Hello,

I am no good at statistics and need help figuring out which statistical test I should use for a hypothetical research project. This is for a paper- I am not actually doing the research. Therefore, I probably have not defined all of variables necessary...but here goes...your help is appreciated!

I am introducing a pilot program on a college campus to help strengthen compliance with a new smoke and tobacco free policy. A survey will be sent out to a random sample of students and faculty at the beginning of the fall term. The program itself is based in the Theory of Reasoned Action/Planned Behavior, so, the survey will be measuring its constructs: attitude, subjective norms, and perceived control.

Next, the pilot program will take place for 1 term, then another survey will be sent out (to a different random sample or should it be sent to the same people?). The three constructs will be measured again.

I want to find out if the program was effective in influencing students attitude, subjective norms, and perceived control.

I think this means I have one independent variable (the program itself) and three dependent variables (the 3 constructs). I also think this means the data is unpaired.

So....what test should I use to compare pre/post program data to see if the program was influential? I'm thinking ANOVA....but maybe Wilcoxon Rank Sum Test?

Help my brain understand why this test is best!

THANK YOU!!!!!
 
#2
what format will you collect data in, i.e. likert scale to judge how they feel about it? if you want to measure how effective, shouldt you do a survey before and after and then test to see if their feelings/ answers/overall mean has changed?
 

Karabiner

TS Contributor
#3
Next, the pilot program will take place for 1 term, then another survey will be sent out (to a different random sample or should it be sent to the same people?).
It would be preferable if it were the same subjects, since within-subject designs
reduce some of the noise which occurs between subjects. But ususally there will
be many subjects which only respond to the first or only to the second survey, and
will therefore have to be excluded from analysis. So you should use the same-
subjects approach only if you have measures at hand which keep attrition low.

The statistical analysis will then of course depend on which approach you
use (same subjects or different subjects).

With kind regards

K.