Using likert scale data and nominal, ratio scale data for Partial Correlation

[sachitha]

Hi All,

If Pearson correlation is used through Partial Correlation (on SPSS) for measuring variables measured on a Likert Scale, can we use specify nominal variables (e.g. Job Role) or ratio variables as for the control variables?

Note: the dependent and the Independent variables will still be on likert scale in this case.


Cheers,
Sach

[CowboyBear]

Hi Sach,

As I understand it, the assumptions that are required for Pearson's correlation (normality, linearity of relationships, continuous data) apply to all variables/pairs of variables used in the partial correlations analysis (including control variables). So analysing Likert scale variables is dubious but potentially justifiable (are you analysing single items or summated scales?), and analysing nominal variables is probably not ideal. Whether the ratio variables are suitable will depend on whether these are continuous normally distributed variables with linear relationships with the other variables.

[sachitha]

Hi,
Thanks for your feedback there.. So as per the likert scales, i'm using the cumulative average for all the questions pertaining to my variables & variable dimensions. Hence they will be continuous.
But now if i'm to use some nominal variables as the controlling variables, what are my alternatives? What more should i do/satisfy to still to Partial Correlation analysis?

Thanks,
Sach.

[CowboyBear]

Hi again Sach,

I'm afraid I don't really know what a specific alternative would be within the partial correlation framework. Maybe someone else has some suggestions.

This page has a nice discussion of partial correlation in general:
http://faculty.chass.ncsu.edu/garson/PA765/partialr.htm