61993_en-US

liliconcom

Correlation analysis using multivariate "BY" vs. splitting dataset

Hello,

I've conducted an experiment pulling samples across a sheet of dough. In my dataset, I have created a dummy variable "belt position" - operator side, middle, non-operator side - for the analysis. I wanted to check correlations for the three separate positions using the multivariate analysis and used the dummy variables in the "by" section. However, if I split the dataset into three separate datasets, the correlations are different than the multivariate analysis. Why is this the case, and is there a incorrect and correct method of comparing the correlations? Thanks!

 

Best,

Eliot

Re: Correlation analysis using multivariate "BY" vs. splitting dataset

Hmmm...I presume you are using the Multivariate platform to evaluate the degree of correlation among the variables? If that's the case, then using a 'by' variable shouldn't change the correlation among the variables for each location. Try using Bigclass.jmp and replicate the process for height and weight by sex and you should see the same correlation for each sex. See attached journals for a 'by' and male only analysis.