[R] error in "predict.gam" used with "bam"

[julian...@elitepartner.de]

Hello everyone.



I am doing a logistic gam (package mgcv) on a pretty large dataframe
(130.000 cases with 100 variables).

Because of that, the gam is fitted on a random subset of 10000. Now when I
want to predict the values for the rest of the data, I get the following
error:





> gam.basis_alleakti.1.pr=predict(gam.basis_alleakti.1,

+
newdata=activisale_join[gam.basis_alleakti.1.complete_cases,all.vars(gam.b
asis_alleakti.1.formula)],type="response")

Error in predict.gam(gam.basis_alleakti.1, newdata =
activisale_join[gam.basis_alleakti.1.complete_cases, :

number of items to replace is not a multiple of replacement length





The following is the code:

#formula with some factors and a lot of variables to be fitted

gam.basis_alleakti.1.formula=as.formula( paste("verl�ngerung ~�,

paste( names(activisale_join)[c(2:10)], collapse="+"), ##factors


paste("s(",names(activisale_join)[c(17,19:29,31:42,44)],")",
collapse="+")) # numeric variables, all count data

)



# complete cases

gam.basis_alleakti.1.complete_cases =
complete.cases(activisale_join[,all.vars(gam.basis_alleakti.1.formula) ])



# modell fitting works on random subset

gam.basis_alleakti.1=bam(gam.basis_alleakti.1.formula,

data = activisale_join[subset.10000, ], family=
"binomial")



# error, no idea why

gam.basis_alleakti.1.pr=predict(gam.basis_alleakti.1,
newdata=activisale_join[gam.basis_alleakti.1.complete_cases,
],type="response")





the prediction on the same subset (subset.10000) works.





It could be that this error is somewhat similar to that described as
sidequestion in

http://r.789695.n4.nabble.com/gamm-tensor-product-and-interaction-td452618
8.html, where simon answered the following:



�> Here is the error message I obtain:
>
vis.gam(gm1$gam,plot.type="contour",n.grid=200,color="heat",zlim=c(0,4))
> Error in predict.gam(x, newdata = newd, se.fit = TRUE, type = type) :
number of items to replace is not a multiple of replacement length
- hmm, possibly a bug. I'll look into it.

best,
Simon�



All the best



Julian



Ps.: > version
_
platform x86_64-w64-mingw32
arch x86_64
os mingw32
system x86_64, mingw32
status
major 3
minor 0.1
year 2013
month 05
day 16
svn rev 62743
language R
version.string R version 3.0.1 (2013-05-16)
nickname Good Sport



package mgcv version 1.7-22




[[alternative HTML version deleted]]

[Simon Wood]

Hi Julian,

Any chance you could send me (offline) a short version of your data,
which reproduces the problem? I can't reproduce it in a quick attempt
(but it is quite puzzling, given that bam calls predict.gam internally
in pretty much the same way that you are doing here).

btw (and nothing to do with the error) given that you are using R 3.0.1
it's a good idea to upgrade to mgcv_1.7-23 or above, for the following
reason (taken from the mgcv changeLog)

1.7-23
------

*** Fix of severe bug introduced with R 2.15.2 LAPACK change. The
shipped version of dsyevr can fail to produce orthogonal eigenvectors
when uplo='U' (upper triangle of symmetric matrix used), as opposed to
'L'. This led to a substantial number of gam smoothing parameter
estimation convergence failures, as the key stabilizing
re-parameterization was substantially degraded. The issue did not affect
gaussian additive models with GCV model selection. Other models could
fail to converge any further as soon as any smoothing parameter became
`large', as happens when a smooth is estimated as a straight line.
check.gam reported the lack of full convergence, but the issue could
also generate complete fit failures. Picked up late as full test suite
had only been run on R > 2.15.1 with an external LAPACK.

best,
Simon

On 08/07/13 10:02, julian...@elitepartner.de wrote:
> Hello everyone.
>
>
>
> I am doing a logistic gam (package mgcv) on a pretty large dataframe
> (130.000 cases with 100 variables).
>
> Because of that, the gam is fitted on a random subset of 10000. Now when I
> want to predict the values for the rest of the data, I get the following
> error:
>
>
>
>
>
>> gam.basis_alleakti.1.pr=predict(gam.basis_alleakti.1,
>
> +
> newdata=activisale_join[gam.basis_alleakti.1.complete_cases,all.vars(gam.b
> asis_alleakti.1.formula)],type="response")
>
> Error in predict.gam(gam.basis_alleakti.1, newdata =
> activisale_join[gam.basis_alleakti.1.complete_cases, :
>
> number of items to replace is not a multiple of replacement length
>
>
>
>
>
> The following is the code:
>
> #formula with some factors and a lot of variables to be fitted
>

> gam.basis_alleakti.1.formula=as.formula( paste("verlängerung ~“,

> ______________________________________________
> R-h...@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.
>


--
Simon Wood, Mathematical Science, University of Bath BA2 7AY UK
+44 (0)1225 386603 http://people.bath.ac.uk/sw283

______________________________________________
R-h...@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.

[julian...@elitepartner.de]

Solved it!! ;)

The problem was that the test-data contained factor-levels the
training-data didn't.
So when trying to fit this new factor-levels to a model which didn't have
this levels, the error occurred.

When excluding the factor-levels not used when fitting the model or when
taking care all levels are used for modell-fitting, everything is fine.

All the best

Julian

-----Ursprüngliche Nachricht-----
Von: Simon Wood [mailto:s.w...@bath.ac.uk]
Gesendet: Dienstag, 9. Juli 2013 09:07
An: julian...@elitepartner.de
Cc: r-h...@r-project.org
Betreff: Re: [R] error in "predict.gam" used with "bam"