[R] How to specify a quadratic term in the contrast function

[Kevin Thorpe]

I don’t know if this will make a difference, but I believe it is generally preferred to use the poly() function (for lm()/glm()) or the pol() function (for rms functions) to include the quadratic (and higher order terms).

Kevin


> On Feb 18, 2025, at 12:23 PM, Ben Bolker <bbolker using gmail.com> wrote:
> 
>  Looking at the 'contrasts' function, it seems to be imported from 'rms' and to take only 'rms'-type fits (which are different from base-R [g]lm() fits) as input.
> 
>  I'm not sure exactly what contrasts you're trying to generate, but I would probably use the emmeans package for the task, since it handles a wide range of model types.
> 
> Alternately, you could use one of the modeling functions from rms, which should work with rms::contrast(), although I ran into trouble for some reason:
> 
> fit1 <- ols(y ~ x + I(x^2), data = mydata)
> 
> > Error in if (!length(fname) || !any(fname == zname)) { :
>  missing value where TRUE/FALSE needed
> 
> I guess rms handles I()-protected terms differently? (The model with just y~x didn't throw an error ...)
> 
> On 2025-02-18 12:00 p.m., Sorkin, John wrote:
>> I am using the contrast package to produce contrasts from a glm model that contains a quadratic term:
>> fit0 <- glm(y ~ x + I(x*x),data=mydata)
>> I would like to know how to specify the quadratic term, I(x*x), in the contrast function.
>> When I try to use
>> contrast(fit0,list(x=4))
>> I receive an error message;
>> #Error in generateData(fit = list(coefficients = c(`(Intercept)` = 0.5262645536229,  :not enough factors
>> When I try to use
>> contrast(fit0,list(x=4,I(x*x)=16))
>> I receive an error message:
>> #Error: unexpected '=' in "contrast(fit0,list(x=4,I(x*x)="
>> Other variants of the call to contrast including
>> contrast(fit0,list(x=4,"x*x"=16)) and
>> contrast(fit0,list(x=4,"I(x*x}"=16))
>> produce similar error messsages.
>> Please see my code below.
>> if (!require(contrast)) install.packages("contrast")
>> library(contrast)
>> help(contrast)
>> x <- 1:100
>> y <- x + x^2 + rnorm(1)
>> mydata <- data.frame(x=x,y=y)
>> fit0 <- glm(y ~ x + I(x*x),data=mydata)
>> summary(fit0)
>> contrast(fit0,list(x=4))
>> #Error in generateData(fit = list(coefficients = c(`(Intercept)` = #0.5262645536229,  :  not enough factors
>> contrast(fit0,list(x=4,I(x*x)=16))
>> #Error: unexpected '=' in "contrast(fit0,list(x=4,I(x*x)="
>> Thank you,
>> John
>> ______________________________________________
>> R-help using r-project.org mailing list -- To UNSUBSCRIBE and more, see
>> https://stat.ethz.ch/mailman/listinfo/r-help
>> PLEASE do read the posting guide https://www.R-project.org/posting-guide.html
>> and provide commented, minimal, self-contained, reproducible code.
> 
> -- 
> Dr. Benjamin Bolker
> Professor, Mathematics & Statistics and Biology, McMaster University
> Director, School of Computational Science and Engineering
> > E-mail is sent at my convenience; I don't expect replies outside of working hours.
> 
> ______________________________________________
> R-help using r-project.org mailing list -- To UNSUBSCRIBE and more, see
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide https://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.


-- 
Kevin E. Thorpe
Associate Professor, Dalla Lana School of Public Health
University of Toronto, Tel: 416-946-8083
Scientific Lead, Clinical Trials Design and Analysis, Hospital for Sick Children