[R] [FORGED] Re: Help to solve modeling problem with gamm

Rolf Turner r@turner @end|ng |rom @uck|@nd@@c@nz
Fri Feb 28 23:16:50 CET 2020

On 29/02/20 8:11 am, Abby Spurdle wrote:

>> There was clearly a non-linear growth pattern such that an additive
>> mixed effects model was proposed to model the behavior of biomass
>> as a function of time and treatments.
> The presence of non-linearity (in x) does not necessarily mean that 
> you can't use a "linear model". In general, linear models work well
> for biological growth curves with small to medium sized data.
> Technically, this is off topic, however, the notion of what is
> linear and what is not linear, is a recurring theme on this mailing
> list.
>> When plotting the residuals a clear cone-shaped pattern was
>> observed
> Based on classical statistics, the obvious approach is to transform
> your data. This is easy to do, however, some expertise is required
> to back-transform parameters/estimates and interpret them.

Although it is indeed off topic I would like to emphasise a point that a 
lot of people don't seem to get.  A linear model is

     *linear in the parameters*

not in the predictors.  For example

     y = beta_0 + beta_1 * x + beta_2 * x^2 + E

is a linear model.  It is linear in (beta_0,beta_1,beta_2).  It is not 
of course "linear in x".


Rolf Turner

Honorary Research Fellow
Department of Statistics
University of Auckland
Phone: +64-9-373-7599 ext. 88276

More information about the R-help mailing list