Help please - what tests do I need to do?

#1
Hi,

I'm struggling to figure out which tests I need to do for my project. I have two things I need to do.

1) I need to know whether males and females varied on a continuous outcome (how strongly the believed a witness was credible in a trial). How would I do that?

2) what analysis should I do - my supervisor has told me to do 'a larger analysis' to see whether condition (male or female) interacted with another variable (continuous) with respect to credibility of the witness.

Any help or guidance would be much appreciated.

Thanks,

Bob
 

Karabiner

TS Contributor
#2
1) I need to know whether males and females varied on a continuous outcome (how strongly the believed a witness was credible in a trial). How would I do that?
t-Test for independent groups. Or, if your sample is quite small, Mann-Whitney U-test.
2) what analysis should I do - my supervisor has told me to do 'a larger analysis' to see whether condition (male or female) interacted with another variable (continuous) with respect to credibility of the witness.
You could perfom a multiple linear regression with 3 predictors: sex (coded 0/1), other_variable, and the interaction between sex and other_variable (which is represented by the product sex*other_variable).

With kind regards

K
 
#4
t-Test for independent groups. Or, if your sample is quite small, Mann-Whitney U-test.
The t-test is also a valid small sample test. It is valid also when the sample size is n=2 in each group. (This is provided that the assumptions are fulfilled. But the Mann-Whitney has also its assumptions and that can mislead people too.)
 

Karabiner

TS Contributor
#5
yup. I was lazy - wanted to prevent myself from discussions about assumptions of the t-test with an OP who doesn't know t-test. Of course you are right. But hardly ever one may see a small-sample analysis where the normality assumption can be relied upon as being fulfilled.

With kind regards

Karabiner