[R] problem with 'gnls'
Bruno Tona
tona at stat.math.ethz.ch
Wed Jan 8 15:12:03 CET 2003
I am working with data measured in a tunnel to estimate the
emission factor of heavy & light vehicles.
I tried to use 'gnls' and I get the following Error:
>> Error in "coef<-.corARMA"(*tmp*, value = c(174.267493382104, 173.412740072763
>> Coefficient matrix not invertible
Here is my R-code: (You can paste it directly into R)
data <- read.table('http://www.stat.math.ethz.ch/~tona/Data/d.plabutsch.neu')
# calculating the starting coeficients:
library("MASS")
library("lqs")
r.CO2.rlm <- rlm(EF.CO2 ~ pLKW.total + Sat + Sun + Fz.total,
data = data, method = "MM", na.action = na.exclude)
START <- coef(r.CO2.rlm); names(START) <- letters[1:5]; START
## a b c d e
## 192.6353049 796.3362315 -35.3434214 -58.8489047 -0.1085019
# 'gnls':
library("nlme")
r.CO2.gnls <- gnls(log(EF.CO2) ~ log(a + b * pLKW.total) + c * Sat +
d * Sun + e * Fz.total,start = START, data = data,
correlation=corARMA(p=2,q=0), na.action = na.exclude)
## Error in "coef<-.corARMA"(*tmp*, value = c(174.267494204742, 173.41274089312
## Coefficient matrix not invertible
I also tried with an AR(3) and an AR(1). The AR(3) gives the same
error like above and the AR(1) gives the error:
>> Error in optim(fn = function(gnlsPars) -logLik(gnlsSt, gnlsPars), par = c(coef(gnlsSt)), :
>> initial value in vmmin is not finite
>> In addition: There were 14 warnings (use warnings() to see them)
Thanks for any help
regards
Bruno Tona
More information about the R-help
mailing list