Write a Reply

low CRP levels in CFS dataset

Dear sleepdata community,

I find it a bit weird that a significant amount (47 subjects) of 'Log value: C-reactive protein levels values' are equal to -2.590267. (https://sleepdata.org/datasets/cfs/variables/lncrppm) Are they valid CRP readings or errors? Should I exclude those subjects from my research?

Thanks,

Hao He

Looking at the original CRP variables, my guess would be that the values for these subjects measured below the floor (based on the lab assay). It looks like values of 0.08 (halfway between 0 and the floor of 0.15) were imputed for these subjects, which produces the consistent negative value in the log-transformed version of the variable.

I suggest you decide whether you want to include or exclude subjects below the floor (or above the ceiling) in lab measurements. You might work more closely with the original CRP variables:

• https://sleepdata.org/datasets/cfs/variables/crpam
• https://sleepdata.org/datasets/cfs/variables/crpam_orig
• https://sleepdata.org/datasets/cfs/variables/crpam_flag
• https://sleepdata.org/datasets/cfs/variables/crppm
• https://sleepdata.org/datasets/cfs/variables/crppm_orig
• https://sleepdata.org/datasets/cfs/variables/crppm_flag

Thanks for using the site!

Thanks for the prompt answer. It is very helpful!