Loop through columns of outcomes

Hello,

Once again, use lapply.

mlist <- lapply(seq_along(m2), function(i) m2[[i]])
names(mlist) <- paste0("mod", seq_along(mlist))

slist <- lapply(mlist, summary)


plist <- lapply(slist, `[[`, 'p.table')


Hope this helps,

Rui Barradas

Em 12-11-2013 13:28, Kuma Raj escreveu:

Thanks for the script which works perfectly. I am interested to do
model checking and also interested to extract the coefficients for
linear and spline terms. For model checkup I could run this script
which will give different plots to test model fit: gam.check(m2[[1]]).
Thanks to mnel from SO I could also extract the linear terms with the
following script:

m2 <- unlist(m1, recursive = FALSE)   ## unlist

First extract the model elements:

mod1<-m2[[1]]
mod2<-m2[[2]]
mod3<-m2[[3]]
mod4<-m2[[4]]
mod5<-m2[[5]]
mod6<-m2[[6]]

And run the following:

mlist <- list(mod1, mod2, mod3,mod4,mod5,mod6)  ##  Creates a list of
models
names(mlist) <- list("mod1", "mod2", "mod3","mod4","mod5","mod6")

  slist <- lapply(mlist, summary)   ## obtain summaries

plist <- lapply(slist, `[[`, 'p.table')   ## list of the coefficients
linear terms

For 6 models this is relatively easy to do, but how could I shorten
the process if I have large number of models?

Thanks


On 12 November 2013 12:32, Rui Barradas <ruipbarradas at sapo.pt> wrote:

Hello,

Use nested lapply(). Like this:



m1 <- lapply(varlist0,function(v) {
         lapply(outcomes, function(o){
                 f <- sprintf("%s~ s(time,bs='cr',k=200)+s(temp,bs='cr')
+
Lag(%s,0:6)", o, v)

gam(as.formula(f),family=quasipoisson,na.action=na.omit,data=df)
       })})

m1 <- unlist(m1, recursive = FALSE)
m1


Hope this helps,

Rui Barradas


Em 12-11-2013 09:53, Kuma Raj escreveu:

I have asked this question on SO, but it attracted no response, thus I
am
cross- posting it here with the hope that someone would help.

I want to estimate the effect of  pm10 and o3 on three outcome(death,
cvd
and resp). What I want to do is run one model for each of the main
predictors  (pm10 and o3) and each outcome(death, cvd and resp). Thus I
expect to obtain 6 models. The script below works for one outcome
(death)
and I wish to use it for more dependent variables.



library(quantmod)
library(mgcv)
library(dlnm)
df <- chicagoNMMAPS
outcomes<- c("death", "cvd", "resp ")
varlist0 <- c("pm10", "o3")

      m1 <- lapply(varlist0,function(v) {
          f <- sprintf("death~ s(time,bs='cr',k=200)+s(temp,bs='cr') +
Lag(%s,0:6)",v)

gam(as.formula(f),family=quasipoisson,na.action=na.omit,data=df)
        })

Thanks

         [[alternative HTML version deleted]]