2 way or 1 way ANOVA...SPSS help!

SideWinder · 16 Mar 2010 at 16:06

I need a bit of help in deciding what ANOVA to use to analyse my results using SPSS.

For my experiment, i have 4 images designed as so..

Attractive lady in a closed posture,
Attractive lady in an open posture,
Neutral lady in a closed posture,
Neutral lady in an open posture.

I chose my participants one of these images so 15 participants for each image. I am discovering the relationship between posture and attractiveness and i am aiming to find out that by having an open posture, it makes you appear more attractive.

Any ideas?

Iraklis F.C. · 16 Mar 2010 at 16:22

You mean ANOVA charts , statistics ?

SideWinder · 16 Mar 2010 at 16:23

Iraklis F.C. said:
You mean ANOVA charts , statistics ?

Yeah.

D.P. · 16 Mar 2010 at 16:27

Although you have 2 dependent variables (hence a 2-way would be the norm), obviously you know the attractive women is more attractive (I assume you have already tested this statistically). Therefore a 1-way would work to test the null hypothesis.

Unless you are looking for interaction effects.

Anway, I would first make sure your data is normal. I highly doubt it is as I have never come across real data which is. Therefore you need to do some wilcoxon rank-sum or kruskall-wallace tests instead of ANOVA.

platinum87 · 16 Mar 2010 at 16:28

You can do it several ways, depends if you compare each to each other, or if you compare one to the other 3

I cant remember much about spss right now, all i can think off, which i did was paired t-test, but cant figure how exactly that would work

MrMoonX · 16 Mar 2010 at 17:00

cant you do a t-test?, surely its a two way as you have two DVs and sounds like a two tailed hypothesis, did you have the women in your pictures rated by an independent rater otherwise you are introducing a confounding variable in that how do gauge attractive/unattractive.

im by no means an expert but am studying psychology at the moment and have used SPSS too much lol

Skeptic · 16 Mar 2010 at 17:32

D.P. said:
Although you have 2 dependent variables (hence a 2-way would be the norm), obviously you know the attractive women is more attractive (I assume you have already tested this statistically). Therefore a 1-way would work to test the null hypothesis.

Unless you are looking for interaction effects.

Anway, I would first make sure your data is normal. I highly doubt it is as I have never come across real data which is. Therefore you need to do some wilcoxon rank-sum or kruskall-wallace tests instead of ANOVA.

As above, really.

Your tutor should help you with this though.

You could do a number of t-tests, but it wouldn't be the most appropriate way to analyse the data. 2 way ANOVA would probably be the most appropriate way, IMO.

killari · 16 Mar 2010 at 18:05

Factorial ANOVA followed by some post hoc tests (i.e Bonferroni) would be a good route to take. T-tests on their own will not work you will get False positives.

Edit: Don't forget you require homoscedacity as well as the normality assumption mentioned above.