ANOVA confusions- Please Help

#1
Hello, I was wondering if you could please offer any advice as I appear to have got rather muddled.

I am writing a resesarch proposal, and I thought I was going to conduct a 2 x 3 mixed ANOVA, followed by a pearson's correlation co-efficient and a multiple regression analysis -and this was okayed by my supervisor. However, I have now read information that has thoughroughly confused me, making me think 1) may need to run a MANOVA (I haven't ever calculated one of these before) and that I may just need to run a regression rather than multiple regresson.

My Variables are as follows
1 between participants independent variable (2 levels dyslexia and no dyslexia)
2 Within participants dependent variables (A- 1 level) (B- 2 Levels x & y).

I have hypothesised that People with dyslexia will report higher A than those without dyslexia.

People with dyslexia will show worse performance in both x and y

There will be negative correlations between A and B (both x and y)

x will be a better predictor of A in both groups
this will effect will be stronger for those people with dyslexia

I thank you in advance for your time. I have tried to work this out but I am getting nowhere- and it doesn't help when someone has said that you are right.

GP x
 
#2
Hello, what do you mean by 1 level and 2 levels in your dependent variables?
Do you mean there are three continuous outcomes, and you only need to know the different between people with and without dyslexia?
 
#3
Thanks Owenpediatricia,

Yes, there are three continuous outcomes although 2 of them (x and y) measure different aspects of the same thing...

I think i'm beginning to see through the fog... as there are three different outcomes (regardless of 2 being associated), would this indicate that my supervisor is indeed correct?

GPx
 
#4
Hello, it depends what your "different aspects of the same thing" means. For the difference in x and difference in y, you might have two outcomes. If you need to know the difference of two groups in the difference between x and y (when x and y tend to highly correlate, I guess it is what you mean by saying 2 levels?), you might need to consider repeated measurement ANOVA.
 
#5
Thanks once more, however, I don't need to know the difference in x and y, but they will (probably) highly correlate. x and y are 2 separate measures of executive ability. Therefore, as I have three within subjects variables and 1 between subjects variable, am I right in thinking I am best off sticking with the mixed model?
 
Last edited:
#6
Hello, What are x and y exactly? To me they do not look like within-subject variables. Different people have different executive ability, and a same person can only have one executive ability. Within-subject variables are variables where a subject have different values under different conditions. The number of conditions is then the number of levels of the variable. But maybe I just didn't get your research question clearly.
 
#7
Hello, to be fair I don't think i've worded it in the best way, I was trying to avoid writing an essay but I think it's ended up just coming out vague :)
I'll list my variables. I have managed to get in a total mess with this one and I really appreciate you taking the time to offer advice.
IV: Dyslexia status
DV 1: Test Anxiety Score
DV 2: Executive function : Verbal Fluency Score, Tower of Hanoi Score

Verbal fluency and Tower of Hanoi both assess different aspects of executive function.

Thank you
 

TheEcologist

Global Moderator
#8
Hello, to be fair I don't think i've worded it in the best way, I was trying to avoid writing an essay but I think it's ended up just coming out vague :)
I'll list my variables. I have managed to get in a total mess with this one and I really appreciate you taking the time to offer advice.
IV: Dyslexia status
DV 1: Test Anxiety Score
DV 2: Executive function : Verbal Fluency Score, Tower of Hanoi Score

Verbal fluency and Tower of Hanoi both assess different aspects of executive function.

Thank you
Factor IV is obviously dichotomous however what are DV1 and DV2?
Will they be continous, discrete, ordinal ect ect
 
#9
Hello my opinion--
I think you only need three ANOVAs (yes, regardless of the correlations). MANOVA might not make many differences.
I don't know what you will do with Pearson correlation or regression, but they are certainly different research questions.