partial or semipartial correlation??

[Hannah Lindsey]

Frank,

I have seen you state that correlational tractography performs a partial Spearman correlation between diffusion and clinical variables, but I'm wondering if it's actually performing a semipartial (aka part) correlation? In other words, are the covariates regressed out of the diffusion metric AND the clinical variable I'm correlating it with, or just the diffusion metric?

Also, when stats are exported from significant tracts loaded onto the connectometry database, they are "raw", so before plotting them, I need to create residuals from them after removing the effects of the covariates, right??

Thanks so much!

Hannah

[Frank Yeh]

Sorry for the late reply because I need to check the source code to confirm.

I have seen you state that correlational tractography performs a partial Spearman correlation between diffusion and clinical variables, but I'm wondering if it's actually performing a semipartial (aka part) correlation? 

In other words, are the covariates regressed out of the diffusion metric AND the clinical variable I'm correlating it with, or just the diffusion metric?

Only the diffusion metric using linear regression.

The steps are:

1. compute a linear fitting between all covariates (including the variable of interest and the intercept) and diffusion metrics

2. for each covariates, here excluding the variable of interest and intercept, adjust the diffusion metrics using the coefficients computed from 1.

3. compute the rank-based Spearman correlation between the adjusted diffusion metrics and the variable of interest.

4. compute the t-statistics from the Spearman correlation

5. apply correlational tracking on the t-statistics

note that the above steps will be repeated 4000 iterations for permuted and non-permuted conditions, respectively. For each iteration, the subjects will be resampled with replacement.

 

Also, when stats are exported from significant tracts loaded onto the connectometry database, they are "raw", so before plotting them, I need to create residuals from them after removing the effects of the covariates, right??

You are right.

Frank

 

Thanks so much!

Hannah

--
You received this message because you are subscribed to the Google Groups "DSI Studio" group.
To unsubscribe from this group and stop receiving emails from it, send an email to dsi-studio+...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/dsi-studio/ca8c1f45-5711-4ab2-9560-94cf9df0bd64n%40googlegroups.com.

[Hannah Lindsey]

Frank,

When performing the Spearman correlation, how are ties handled when ranking values? For example, my clinical variables are largely discrete, so there are several instances of multiple subjects with the same ("tied") value. Are these given the average tied rank value or is some other correction for ties applied? I just want to make sure I'm plotting the data correctly! 

Thank you!

[Frank Yeh]

I checked the code. The computation does not consider or handle tie conditions.
I am listing this as the to-do list on the next release.
Frank

> To view this discussion visit https://groups.google.com/d/msgid/dsi-studio/f52ee83b-4956-49d6-97ed-a0ce9feb5e12n%40googlegroups.com.

[Hannah Lindsey]

Thank you!

_________________________________

Hannah M. Lindsey, PhD

Research Associate

TBI and Concussion Center

Department of Neurology

University of Utah

On Jan 28, 2025, at 2:13 pm, Frank Yeh <fran...@gmail.com> wrote:

I checked the code. The computation does not consider or handle tie conditions.