In R version 3.6.3, my predict.gam function gives a result. When I use R 4.x, I get errors for some types of predict, but not others.
My goal is to be able to use this code in R4.
library(gam)
gm1 <- predCol ~ s(VCF_nontree) + s(VCF_nonvegt) + s(VCF_treecov) +
s(CCI_URBAN) + s(GPW_POP)
gm2 <- predCol ~ offset(offset) + s(ACC) + s(EARS) + s(MAT) + s(PTA) +
s(SLP) + s(SOC) + s(TWI)
set.seed(42)
c2 <- data.frame(
predCol = abs(rnorm(677072, 0.022, 0.03)),
VCF_nontree = rnorm(677072, 6533, 16.3),
VCF_nonvegt =rnorm(677072, 15.9, 12.03),
VCF_treecov = rnorm(677072, 18.1, 18.3),
CCI_URBAN = rnorm(677072, 0.00412, 0.013),
GPW_POP = rnorm(677072, 2.23, 1.28),
offset = rep(NA, 677072),
ACC = rnorm(677072, 612.6, 456.5),
EARS = rnorm(677072, 731.88, 291.4),
MAT = rnorm(677072, 22.69, 2.56),
PTA = rnorm(677072, 124.13, 307.9),
SLP = rnorm(677072, 2.41, 3.71),
SOC = rnorm(677072, 0.609, 0.33),
TWI = rnorm(677072, 118.2, 15.35)
)
fitModel1.1 <- gam::gam(gm1,
data = c2,
family = quasibinomial)
c2$offset <- as.numeric(predict(fitModel1.1,
type = "link"))
fitModel1 <- gam::gam(gm2,
data = c2,
family=quasibinomial)
fm1 <- predict(fitModel1,
se.fit = TRUE,
type = "response")
in R3 I get a result, in R4 I get this error:
Error in NextMethod("predict") : no method to invoke
R3, gam 1.20, R4 gam 1.22-2
in R4, this call to predict works:
fm2 <- predict(fitModel1,
se.fit = TRUE,
type = "link")
update the package maintainer has fixed the issue, and new version of gam on cran no longer need a workaround, and work as designed!
Following the advice of @GavinSimpson, I made a workaround: a custom predict.gam() function that can be called in an Rscript using
source("custom_predict_Gam.R") I save the below script in my working directory, then in R 4, use source(), then call the function like this:
fm1 <- custom_predict_Gam(fitModel1,
se.fit = TRUE,
type = "response")
and get results identical to those from R 3.
Thanks for everyone's input. This has made me happy. gam::happy...
On the off chance that this is useful to anyone else, here is the function that I save as custom_predict_Gam.R:
# custom_predict_Gam.R
# this file replaces the predict.gam on line 18 with predict. The reason is because predict.gam is broken in R 4.
custom_predict_Gam <- function (object, newdata, type = c("link", "response", "terms"),
dispersion = NULL, se.fit = FALSE, na.action = na.pass, terms = labels(object),
...)
{
type <- match.arg(type)
if (missing(newdata)) {
if (inherits(object, "Gam") && !is.null(object$smooth)) {
if (se.fit)
switch(type,
response = {
out <- predict(object, type = "link", se.fit = TRUE, ...)
famob <- family(object)
out$se.fit <- drop(out$se.fit * abs(famob$mu.eta(out$fit)))
out$fit <- fitted(object)
out
},
link = {
out <- NextMethod("predict")
out$fit <- object$additive.predictors
TS <- out$residual.scale^2
TT <- ncol(object$var)
out$se.fit <- sqrt(out$se.fit^2 + TS * object$var %*%
rep(1, TT))
out
},
terms = {
out <- NextMethod("predict")
TT <- dimnames(s <- object$smooth)[[2]]
TT = intersect(terms, TT)
out$fit[, TT] <- out$fit[, TT] + s[, TT]
TS <- out$residual.scale^2
out$se.fit[, TT] <- sqrt(out$se.fit[, TT]^2 +
TS * object$var[, TT])
out
})
else switch(type,
terms = {
out <- NextMethod("predict")
TT <- dimnames(s <- object$smooth)[[2]]
TT = intersect(terms, TT)
out[, TT] <- out[, TT] + s[, TT]
out
},
link = object$additive.predictors,
response = object$fitted)
}
else {
if (inherits(object, "Gam")) {
if (type == "link" && !se.fit)
object$additive.predictors
else NextMethod("predict")
}
else UseMethod("predict")
}
}
else newdata.predict.Gam(object, newdata, type, dispersion,
se.fit, na.action, terms, ...)
}