Hi, I had this kind of model:
lmer(RunningDistance ~group+sex+condition+(1|id)+(1|day)+(1|sprint), data=data)
But, the Running Distance is a truncated variable, because individuals just had a racetrack of 1m to run, not more. In this model even the residuals seem to be truncated.
I know in this cases the coefficients may be severely biased. In this case it will be less than if I had a untruncated variable. The truncreg function (truncreg package) seems to be ideal, but it not take into account the random variables.
I really appreciate any suggestion!
Thanks!
Is this truncated or censored? I think you have censored data.
Regardless, the only way I know to do is through a Bayesian approach using a package like brms. It’s actually pretty easy to use.
https://www.jstatsoft.org/article/view/v080i01/v80i01.pdf
You may find some help here:
https://goo.gl/tBuccU
Article Mixed model analysis of censored longitudinal data with flex...
An this question was also discussed in the R mailing list:
https://stat.ethz.ch/pipermail/r-help/2007-April/130376.html
I am somewhat confused. If the dependent variable is distance, and the distance is always 1 mile, then the model should not work because there is no variability in the dependent variable. I suppose stressed individuals might not be able to make a mile, especially if they are given a fixed time to complete the task. Maybe they are physically challenged in some way (overweight, cardiac issues, disabled). Without some other conditions, I would expect most individuals could traverse 1 mile given sufficient time. If a small fraction of people were able to traverse the 1 mile track, then I can see this as censored data. If everyone completed the track, then I have trouble seeing how this approach will work.
Given that all individuals are treated equally, why can you not restrict discussion to a 1 mile sprint? Without data, it would be a bad idea to extrapolate to a 50 mile marathon. I would think it is also unsuitable if you are trying to predict the maximum distance someone can run.
If you have a video of the sprint, could you reanalyze the data at 100 feet, 200 feet, ... 5280 feet? Is a pattern visible?
@Timothy, would it help knowing that Solimary is working with arthropodes? ;)
Thank you Jochen. I didn't look at the bio :(
1) If given sufficient time, all scorpions will traverse the 1 meter. If all (or most) of the scorpions make the 1-meter mark, then there is too much "missing data." One either needs to start over with a larger arena, or redefine the problem. Talk about initial movement (within the first meter), and say nothing about long term movement (>1 meter).
2) If there is video of the scorpions, then can you reanalyze the data and look at an arena of 30, 60, 90, and 100 centimeters? How does the answer change as arena size changes?
- Badges
- Science topic
- Similar topics
- Medicine
- Physiology
- Biosignals
- Biomedical Signal Processing
- Statistical Analysis