1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
|
% $Id: coefFrame.Rd 776 2005-12-01 16:54:02Z nj7w $
%
\name{coefFrame}
\alias{coefFrame}
\title{Return model parameters in a data frame}
\description{Fits a model to each subgroup defined by \code{by}, then
returns a data frame with one row for each fit and one column for each
parameter.
}
\usage{
coefFrame(mod, data, by = NULL, fit.on = TRUE, fitfun, keep.unused.levels = TRUE, byvar.sep = "\001", ...)
}
\arguments{
\item{mod}{a model formula, to be passed to by \code{fitfun}.}
\item{data}{a data frame, row subsets of which will be used as the
\code{data} argument to \code{fitfun}.}
\item{by}{names of columns in \code{x} that will be used to define the
subgroups.}
\item{fit.on}{a logical vector indicating which rows of \code{x} are
to be used to fit the model (like the \code{subset} argument in a
lot of other functions). Can be
given in terms of variables in \code{x}}
\item{fitfun}{a model fitting function (e.g. lm, nls). More
specifically, a function that expects at least a formula object (as the
first argument) and a data.frame object (passed as an argument named
\code{data}) and
returns a model object for which a \code{coef} method
has been defined (e.g. coef.lm, coef.nls) to extract fit values of
model parameters.}
\item{keep.unused.levels}{Include rows in output for all unique values
of \code{by}, even those which were excluded by
\code{fit.on}. The default value \code{TRUE} should be left alone
if you are going to go on to pass the result to \code{backFit}.}
\item{byvar.sep}{passed to frameApply, used to form the
subsets of the data.}
\item{...}{other arguments to pass to \code{fitfun}.}
}
\value{a data frame with a row for each unique row of \code{x[by]}, and
column for each model paramter, as well as columns specified in \code{by}.
}
\examples{
# load example data
library(gtools)
data(ELISA)
# Coefficients for four parameter logistic fits:
coefFrame(log(Signal) ~ SSfpl(log(Concentration), A, B, xmid, scal),
data = ELISA, fitfun = nls,
by = c("PlateDay", "Read"),
fit.on = Description == "Standard" & Concentration != 0)
# Coefficients for linear fits:
coefFrame(log(Signal) ~ log(Concentration),
data = ELISA, fitfun = lm,
by = c("PlateDay", "Read"),
fit.on = Description == "Standard" & Concentration != 0 )
# Example passing arguments to fitfun, and example of
# error handling during model fitting:
ELISA$Signal[1] <- NA
coefFrame(log(Signal) ~ log(Concentration),
data = ELISA, fitfun = lm, na.action = na.fail,
by = c("PlateDay", "Read"),
fit.on = Description == "Standard" & Concentration != 0 )
}
\author{Jim Rogers \email{james.a.rogers@pfizer.com}}
\keyword{models}
|
