[R-sig-ME] Help on random regression model with correlated measures on consecutive days

Manuel Ramon m.ramon.fernandez at gmail.com
Mon Feb 15 08:59:42 CET 2016

Dear list members,
I want to examine the effect of climate on individual behavior (in my case
animal productive performance). My model has different levels of variation:
- 1st: individual production level, being high at the beginning and falling
with time. It is expected that climate effect was higher when individuals
yield at higher levels.
- 2nd: day of climate measurement from day of production measurement. I
have climate measures for the day of production measurement and the
previous 30 days. It is expected that the effect of climate conditions on
the day of production measurement was higher that on the day 15 previous to
production measurement. At this point, it is also important to take into
account that the effect of climate conditions when temperature equals, for
example, to 20ºC changed depending on whether in the days before it was
cold or hot (that is, is not the same 20ºC from 30ºC than from 10ºC).

To address this study, I run a random regression model that included, in
addition to other fixed effects, a term for production (quadratic
polynomial), a term for average temperature on the day of production
measurement and the 3 previous days (quadratic polynomial) and a random
individual effect with nested quadratic polynomials for production and
average temperature.

The model run well, but I think that production and temperature effects are
confounded and that I am not considering the lagged effect of temperature
on days previous to production measurement, especially if previous days
were hotter or colder.

My question is how to design a model that include the production trend, the
effect of temperature depending on production level and how this
temperature effect changes depending on the day in which temperature was

Thanks for your help,


	[[alternative HTML version deleted]]

More information about the R-sig-mixed-models mailing list