1

I have some data I would like to look at to investigate associations between specific TLR genotypes (wild type, heterozygous, homozygous) and both a binary outcome variable (death) and continues (biomarkers).

Would an ordinal logistic regression be appropriate here? Or separate logistic models between the wild type and homozygous, and wild type and heterozygous?

When I look at covariates- do i look at associations with TLR or associations with outcome?

Lucy
  • 189
  • 1
  • 12
  • The death variable doesn't look like ordinal to me. – SmallChess Feb 02 '17 at 23:34
  • what test should i use to account for the allele frequencies? – Lucy Feb 04 '17 at 00:56
  • You have a continuous and a nominal variable. Why can't you just use a simple logistic regression? – SmallChess Feb 04 '17 at 03:03
  • I have this intuition that i need to account for the allele frequencies somehow in my model (whether the subject is wild type vs heterozygous vs homozygous) and explain whether the outcome is associated more so 2 copies of the allele than 1 copy, and more so with 1 copy of the allele than none (wild type). – Lucy Feb 04 '17 at 06:58
  • 1
    Are the continuous biomarkers also to be used as predictors of death? Do you have information about time to death? – EdM Feb 04 '17 at 10:36
  • 2
    Please look at [this thread](http://stats.stackexchange.com/q/5387/28500) for background. "Ordinal logistic regression" typically means having an [ordinal response variable](https://en.m.wikipedia.org/wiki/Ordered_logit); here you have binary outcome and ordinal predictor. – EdM Feb 04 '17 at 10:49
  • Thank you! The continuous biomarkers will be looked at as an outcome variable but also mostly to explain the association of the SNP to the death outcome. I do not have information on time to death. – Lucy Feb 05 '17 at 06:47

0 Answers0