Pre- and post-test comparison, is there a minimum sample size (behaviour change treat

#1
Hi

I have to measure the efficacy of a group based treatment for the perpetrators of domestic violence. I started out trying to organise an rct, but it looks like all i will be able to do is a pre and post-test comparison given the difficulty in establishing a comparable control group. I understand i would have needed around 40-50 participants in control and experimental groups to do the rct to get a valid statistical analysis, is there a similar directive for the pre- and post-test analysis? will i need a 40 participants?

Thanks in advance for your time
 

noetsi

No cake for spunky
#2
Re: Pre- and post-test comparison, is there a minimum sample size (behaviour change t

Is rct random assignment?

The answer (which is a question of statistical power) depends on your specific test, the alpha level you use for your hypothesis, what power you want, and the effect size you assume. Gpower is a good site to actually determine what the number is, but there is no way to provide this answer with the information you have provided.
 

hlsmith

Less is more. Stay pure. Stay poor.
#3
Re: Pre- and post-test comparison, is there a minimum sample size (behaviour change t

You also have to determine which test you will be conducting. You have not described the dependent variable at all, so we are not able to say - you need to do a sample size calculation for a two sample t-test. But perhaps you already knew this and just wanted to see if there was a difference between RCT and non-RCT design in regards to the n and m-values.

The main difference will be in establishing causality and describing your findings.
 

Karabiner

TS Contributor
#4
Re: Pre- and post-test comparison, is there a minimum sample size (behaviour change t

If it's a group based intervention then the additional problem arises
that one cannot analyse data from groups as if group members were
completely independent from each other, i.e. one has to take into
account clustering.

With kind regards

K.
 
#5
Re: Pre- and post-test comparison, is there a minimum sample size (behaviour change t

Thankyou all

that is all useful

are there statistical tools built to deal with clustering effects - in relation to this issue, please note that data on recidivism (which means any display of abusive behaviour) will be collected via means independent of the participants for example via Police records and partner reports as well as from the participants themselves - will that negate clustering effects?
 
#6
Re: Pre- and post-test comparison, is there a minimum sample size (behaviour change t

yes i see, the dependent variables will include measures of behaviour change (in effect the level of abuse they perpetrate against their partner) which will be measured through questionnaires with their partner. Other measurements will include psychometric tests before and after the treatment programme.

"A sample size calculation for a two sample t-test" no i'm just learning about this stuff - i'll go and revise t-tests - i'll be back once i know what you are talking about - before i go, is that basically how i should treat the pre and post test results, much the same as the experimental and control in an experiment?

Interesting that you mention causality there also, i have been trying to find suitable measurements which i can employ throughout the 6 month treatment to help understand what it is within the treatment that makes a difference to different personality types