[R] NLME model question

[Douglas Bates]

Darren Shaw <Darren.Shaw at ed.ac.uk> writes:

> Dear R-helpers
> 
> I have a problem related to the use of NLME
> 
> I think is simply a matter of getting the nlme coding correct, but i
> cannot get my brain around it
> 
> 
> I am analysing some 24 growth curves of some cells , and i wanted to
> say that there are significant differences between the curves in two
> parameters that describe the pattern of growth.  these parameters are
> from a logistic (r & k) .
> 
> 
> i have attempted to construct a self starting routine for nlme ie:
> 
> SSGrowth_function(x, r, k)
> {
>          .expr2 <- (k - 100000)/100000
>          .expr5 <- exp(((r * -1) * x))
>          .expr7 <- 1 + (.expr2 * .expr5)
>          .expr13 <- .expr7^2
>          .value <- k/.expr7
>          .actualArgs <- match.call()[c("r", "k")]
>          if(all(unlist(lapply(as.list(.actualArgs), is.name)))) {
>                  .grad <- array(0, c(length(.value), 2), list(NULL,
>                  c("r", "k")))
> 
>                  .grad[, "r"] <-  - ((k * (.expr2 * (.expr5 * (-1 *
>                  x))))/.expr13)
> 
>                  .grad[, "k"] <- (1/.expr7) - ((k * (1e-005 * .expr5))/.expr13)
>                  dimnames(.grad) <- list(NULL, .actualArgs)
>                  attr(.value, "gradient") <- .grad
>          }
>          .value
> }
> 
> where x = time, 100000 = known starting conditions, r = growth and k =
> carrying capacity
> 
> 
>   i guessed i should then write
> 
> nlme(NoofCells~SSGrowth(Time,r,k),fixed=r+k~1,data=CellData,random=r+k~1)
> 
> 
> This runs and tells me that r & k's do differ
> 
> BUT.  The "CellData" actually consists of replicates - ie there are 4
> cell types, but they are done 6 times each.  Therefore, I do not want
> to ask if there are significant differences in r & k between 24 sets
> of data ("Runs")- rather I want to be able to say that there are
> differences between the four cell types occurring 6 times each.  So
> how do i  incorporate "CellType" explicitly into my model structure??
> 
> 
> i.e. If i was lust looking at say linear growth and was using lme I
> would have written something like
> 
> 
> lme(NoofCells~Time*CellType,random=~1|Runs,data=CellData)
> 
> Any thoughts/suggestions gratefully received

I believe you want

nlme(NoofCells~SSGrowth(Time,r,k),fixed=r+k~1,data=CellData,random=r+k~1|Runs)