Correlation between three platforms

Question

[tdfyoder]

Hello,

I have 40 samples, each sampled was tested with the same 27 analytes on three different platforms. I want to test for correlation between the three platforms in R.

I have used a multiple linear regression lm( y ~ x + z) approach; where Y is the trial data from a new platform and x and z are existing platforms in house. I'm afraid x & z not independent values, same samples being tested- expecting the analyte quantified to be similar. I generated lm models for each of the analytes. To test for collinearity I ran VIF() from the regclass package on the lm, values range from 1-200).

I am wondering if using Pearson's r is the best metric and reviewing pair-wise correlation coefficients?

My assumption thus far has been that linear correlations exist, though surveying scatter plots shows this is not necessarily the case.

Suggestions on how to proceed would be welcomed. Thanks.

-T

Answer 1

[tdfyoder]

I did end up doing a pair-wise analysis between platforms, fitting a linear regression but more importantly utilizing a Bland Altman analysis to quantify bias between the platforms (paper). I did include the multivariable regression w/ vif as evidence of correlation between predicting variables-- though I emphasized the results of the Bland Altman.