# Ineed help with statistics

#### Carrot5

##### New Member
Hi, I need help with my research. I've used three different measures of one construct and my aim was to explore if those three measures are correlated. One of them is a questionnaire, second is a laboratory task where participants had to recognize something from pictures and give a correct answer, and the third is also a laboratory task where participants had to choose how the pictures made them feel. So I conducted correlations and they were not as it was expected, they were not correlated completely. So, my question is what statistical procedures can I use to see if there were problems with my measures. Would Confirmatory factor analysis be useful with the questionnaire, so that I can see if my data really show three different factors as the developers of the measure suggest? For the task with correct responses, I am not sure what to do. But with the third task, I was thinking to use chi-square, since authors of that measure gave means for all assessments of pictures, for future researchers to compare their data with. Am I right, or the procedures I am about to do won't help me? And what procedures can I use if not those?
I am asking this so that I can better explain why my results aren't as expected.

Last edited:

#### Karabiner

##### TS Contributor
One of them is a questionnaire, second is a laboratory task where participants had to recognize something from pictures and give a correct answer, and the third is also a laboratory task where participants had to choose how the pictures made them feel.
So the first variable was measured as a score from a questionnaire. How were the other two variables measured?
So I conducted correlations and they were not as it was expected, they were not correlated completely.
What do you mean by "not correlated completely"? Do you mean the correlation coefficients
were < 1 ? How large were they? And why did you expect the variables to correlate completely?

So, my question is what statistical procedures can I use to see if there were problems with my measures.
What kind of problems do you mean?

With kind regards

Karabiner

Last edited:

#### Carrot5

##### New Member
Thank you for your reply. The second variable was measured as a number of correct identifications of different facial expressions, and in the third were sums of evaluations of pictures in two aspects(I don't know if that is the right word for that, or if I explained it correctly, sorry )

I didn't mean they didn't correlate completely as they didn't correlate r=1, I expressed myself wrong, I'm sorry. What I meant is that the subscales (or components) of the questionnaire did not all correlate with my other two tasks (some of them did). So I thought that I could use some statistical procedures to see why that exactly happened, that's what I meant when I said "problems with my measures"

#### Karabiner

##### TS Contributor
So your questionnaire comprised several scales, and some of them correlated with r=0.00
with the other 2 measurements, other scales showed correlation coefficients r > 0.00 ?

Since neither the technical details of the questionnaire nor the theoretical background is
known, I'd assume that the results are just accepted as they are?

With kind regards

Karabiner

#### Carrot5

##### New Member
Yes, some of them had significant correlations and some of them were not significant.
All three instruments theoretically measure the same construct, that is why they were expected to be correlated. Previous research didn't show consistent results of correlation between that type of measures of that construct, so my aim was to explore whether these specific instruments correlate or not. And I thought that I could use other statistical procedures for a better explanation of these results.
I'm not sure I understand your last question, but I guess yes

#### Karabiner

##### TS Contributor
Yes, some of them had significant correlations and some of them were not significant.
You say yes, but what you then describe is something very different from what I assumed.
"significant" or "not significant" are quite meaningless, at least without description of
sample r's and sample sizes.

All three instruments theoretically measure the same construct, that is why they were expected to be correlated.
You said that you are not happy with some subscales' correlations with
the other measures. But if the subscales/dimensions are orthogonal, then
this is something which can happen all the time.

And I thought that I could use other statistical procedures for a better explanation of these results.
background and the conceptual and technical details of the questionnaire
and its composition.

With kind regards

Karabiner