Hi, I am currently trying to do dose-response curves
using weighted 4-parameter model (4PL). The weighting was based on
1/(expected variance) derived from historical data. I tried both drm() from drc
package, and nls(), found very different
results derived from drm() vs. nls() using "weights=" argument.
d1<-read.table("d1.txt",sep='\t',header=T,row.names=1)
drm(y~x,weights=1/exp((-1.8286)+0.05481*y-0.000405225*(y-48.34177)^2),data =
d1,fct = LL.4(),control=drmc(relTol=1e-12))
b:(Intercept) c:(Intercept) d:(Intercept) e:(Intercept)
-1.5859 0.7823 47.3591 38.7717
nls(y~c+(d-c)/(1+(x/e)^b),weights=1/exp((-1.8286)+0.05481*y-0.000405225*(y-48.34177)^2),data=d1,
start=list(b=-1, c =0, d=100, e=150))
b c d e
-1.0076 -0.1143 104.9470 152.5558
But if I run unweighted dose-response curve, both drm() and nls() gave the same
results:
drm(y~x,data = d1,fct = LL.4(),control=drmc(relTol=1e-12))
b:(Intercept) c:(Intercept) d:(Intercept) e:(Intercept)
-1.1416 0.6986 102.8613 135.9371
nls(y~c+(d-c)/(1+(x/e)^b),data=d1, start=list(b=-1, c =0, d=100, e=150))
Nonlinear regression model
b c d e
-1.1416 0.6987 102.8613 135.9373
Any suggestions would be appreciated!
John
y x
17 2.34401224931698 4.88
18 33.2408581210951 78.13
19 90.2562925955097 1250
20 106.021704365726 20000
21 100.222797235699 5000
22 78.5794363031895 312.5
23 12.7579568230370 19.53
24 1.04435840221690 1.22
______________________________________________
R-help@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.
