Hello, I am struggling to find a suitable statistic test for one of my studies... and hope that anyone can help me with it:
I have 4 different conditions, and I want to test if my results differ on those conditions. When the dependent variable is continuous it is easy to find a statistical test, however when it is a categorical variable I am not finding the right one.

Some details that may help:
Number of users: 23
Categorical Variable: 4 different values (not ordinal.. but can be ordinal if that's the only way)

Any suggestions?

Thanks for the help in advance,
João Guerreiro

ok so if you want to see if there are significant differences between 4 conditions, where the same participants are in every condition, then you could use a repeated measures ANOVA or a MANOVA

Thank you for your answer. I think ANOVA can only be used for continuous data, and my dependent variable is categorical instead. Am I right?

Thanks

Is it categorical data within 4 conditions? or do you just mean that you have 4 conditions and therefore it is categorica?

if its four different conditions then you can use repeated measures ANOVA, but it maybe would be better to use MANOVA, with each condition as a dependent variable.

It has 4 different conditions (the repeated measures part).
Besides that, the dependent variable is categorical. I'll say the possible values to better comprehension:
0- Did not mention
1- Did not know
2- Mentioned Induced
3- Mentioned Spontaneously

Since the variable is categorical instead of continuous, I think it is incorrect to use the ANOVA. I guess I could use the Friedman Test (the non-parametric for the repeated measures ANOVA) if I consider the variable ordinal; however, I think it is more correct to consider it nominal instead.

======= Date Modified 12 Sep 2011 14:52:17 =======
I believe that Pearson's chi square test is for categorical data.... But i may be being a bit thick. I'm sleepy.

yeah you can use chi-square for categorical data like that, although maybe you could use logistic regression, using your 4 conditions as predictors and having a dichotomous outcome where ou collapse the first two answers (i.e. not mentioned) as 0 and the last two (i.e. mentioned) to 1. But that would depend on your study. and with just 23 participants you don't really have the power so chi-square is probably your best bet.

I also thought about using chi-square, but is that suitable for within-subjects/repeated-measures? I read that one of the assumptions is independent observations (each person or case can be counted only once)

Hmz. I missed the within subjects bit... are the questions related? So... say you are looking into whether someone likes chocolate... they will only fit into one of your categories. If there are no other responses that are related to chocolate, say the next thing you are considering is if they like the colour yellow, then the statements are sufficiently independent that i think a chi-square will be acceptable.(But i would want someone cleverer than me to verify this!)

If, however, your statements are related such that you are considering if the participant liked chocolate and whether they then go onto mention galaxy and dairy milk as types of chocolate then your responses would not be independent so chi-square is not applicable and i'm afraid i don't know an alternative off the top of my head.

Unless... there might be a within-subjects variant that i am unaware of...

(My stats is a bit rusty so someone else may need to verify this....)

I'm not best at stats but I'm confused about your variables. Are you trying to predict something? Or just comparing four variables? I think you can use chi-square if you're comparing yes... but with 23 participants and the number of variables I'm not sure the power will be that great to be honest.

Yes I am comparing. I've found something that seems very similar to what I want, but that only allow 2 categories (for yes/no variables). That's the Cochran Q test. If it would allow more than 2 categories I guess it would be it...

The penny droppedl I'm sorry then, you're right, you can't use chi-square...

I'm not familiar with the Cochran Q test so can't comment much on it apart from I've had a look at wikipedia and it looks like it's what you need. Maybe someone else will be able to help?

have you looked at log-linear analysis? that may help with the independence assumption, but I know little about it.