2

My objective is to see if there is a significant difference in BHB concentration between age categories in farm animals. Farm should be a random effect in the model. The issue is that BHB concentration is left-truncated (lower detection limit at 96). I have 94 data rows, of which 63 are at the lower limit of detection. Here is the distribution for the whole dataset and the distribution in each age category:

enter image description here

enter image description here

If BHB had a normal distribution, I would just use a linear mixed model like this and do a post-hoc test like Tukey for pairwise comparisons between age categories:

lmer(BHB ~ age_categories + (1|farm), data=file)

Is there a package in R to do a similar mixed model but adapted to censored data? I think of VGAM or censReg but I'm stuck when I need to write the equation as well as to calculate the pairwise comparisons to see if there are significant differences in BHB concentration between age categories.

Any help is appreciated.

Falco
  • 139
  • 10
  • Not exactly censored data, but you could use quantile regression. R has the lqmm package which could help. You also have very skewed results, and quantile reg is OK with that.

    As an aside, do you really need a test? Herman Friedman, who was my favorite professor in grad school, would say that this passes the IOTT -- the interocular trauma test -- it hits you between the eyes.

    See this thread for more on the IOTThttps://stats.stackexchange.com/questions/458069/source-for-inter-ocular-trauma-test-for-significance

    But maybe you need a test to satisfy an editor or something.

     – Peter Flom Jun 26 '23 at 10:21
  • Thank you very @PeterFlom, I will have a look at quantile regression. I need a test because I need to publish the results. I have many non-truncated variables in my dataset (for which I use a simple mixed model) and BHB is the only truncated variable. I need a method that is consistent with the mixed models I use for the other variables, but adapted to truncated data. – Falco Jun 26 '23 at 11:02

0 Answers0