Correlation

Avatar for sneaks

I need to do a spearman's correlation, as I have gender (0, 1) to add into the mix.

However, my sup reckons you can report (in the SAME table) pearsons and spearmans. So just reporting spearmans coefficients for the ones related to gender and then the rest as pearsons.

Is this right? it sounds a bit dodgy to me

D

I have presented spearman or pearsons statistical data in the same table but used one or the other test depending on parametric or non-parametric distribution of the data for a given parameter. I have not specified in the data table which test was used, so I guess I will find out later in my viva if this poses a problem! I have explained this in my data handling section and I have not had any queries from my supervisory team.

Avatar for sneaks

Ok, so you think its fine to put e.g. all the correlations relating to gender as spearman's coefficients in the same table as pearsons coefficients for all the other variables?

Ok - I may put a footnote to say it. Plus I will put (see Dunni, 2011)

D

The two tests are essentially used to test curvilinear correlations but one is more accurate in non-parametric, the other for parametric distributions. It makes sense that the appropriate test is used for the data distribution ie gender data is most likely normally distributed. I checked my data distribution using SPSS and then took the appropriate statistical info. I think it wise to footnote the test used to add clarity and this is something I may need to add post viva (but I will wait and see!). Perhaps present some of the data to your supervisor to clarify before doing them all this way.

Just to add, in some of my data summary tables the data may show rho, t or R2 values depending on the data trend, but again this is detailed in my data handling section.

17932