R help - May 2014 - Lack-of-fit test for linear mixed effects model (lme) including two fixed effects

Dear list members,

   I fitted a linear mixed effects model using the lme function from nlme
   package. In the model I included two fixed effects, one being continuous and
   one a factor having 4 categories. Furthermore, I have one random effect (id)
   which I want to include as a random intercept only. I used the following
   code:

   fit1 <- lme(outcome ~ fixed1 + fixed2, random = ~1|id)

   Now I would like to perform a lack-of-fit test. In a previous post (where
   there  was only one continuous fixed effect) I have seen the following
   suggestion:

   fit1 <- lme(outcome ~ fixed1, random = ~1|id, method="ML")
   fit2  <-  lme(outcome  ~ factor(fixed1, ordered=TRUE), random = ~1|id,
   method="ML")
   anova(fit1,fit2)

   Now my questions are:
   1) How do I perform a lack-of-fit test with one continuous and one factor as
   fixed affects?
   2) Is it necessary to set method=ML for the lack-of-fit test?
   3)  If  I have to use method=ML for the lack-of-fit test, should I use
   method=ML also in my model taht I would like to interpret? In my original
   model I used REML, but only because this is the default in lme and I did not
   change it.

   As I have not applied a lack-of-fit test before, I would really be glad for
   any help!

   Best regards,
   John

This is really the sort of issue that should be discussed with a local
statistical expert.

(1) It is,  a statistical, not an R programming question;

(2) Would be better posed on the R-sig-mixed-moodels list anyway;

(3) As lack of fit tests tend to be basically useless and
scientifically meaningless, the real question is what is the context
for your question -- what are the meaningful practical  (not
statistical) concerns that you want to address with the test? Again,
the sort of thing best discussed in person with a local expert.

Cheers,
Bert


Bert Gunter
Genentech Nonclinical Biostatistics
(650) 467-7374

"Data is not information. Information is not knowledge. And knowledge
is certainly not wisdom."
H. Gilbert Welch




On Sat, May 17, 2014 at 10:22 AM, Isabella Zwiener <izwiener at gmx.de>
wrote:
>
>    Dear list members,
>
>    I fitted a linear mixed effects model using the lme function from nlme
>    package. In the model I included two fixed effects, one being continuous
and
>    one a factor having 4 categories. Furthermore, I have one random effect
(id)
>    which I want to include as a random intercept only. I used the following
>    code:
>
>    fit1 <- lme(outcome ~ fixed1 + fixed2, random = ~1|id)
>
>    Now I would like to perform a lack-of-fit test. In a previous post
(where
>    there  was only one continuous fixed effect) I have seen the following
>    suggestion:
>
>    fit1 <- lme(outcome ~ fixed1, random = ~1|id, method="ML")
>    fit2  <-  lme(outcome  ~ factor(fixed1, ordered=TRUE), random =
~1|id,
>    method="ML")
>    anova(fit1,fit2)
>
>    Now my questions are:
>    1) How do I perform a lack-of-fit test with one continuous and one
factor as
>    fixed affects?
>    2) Is it necessary to set method=ML for the lack-of-fit test?
>    3)  If  I have to use method=ML for the lack-of-fit test, should I use
>    method=ML also in my model taht I would like to interpret? In my
original
>    model I used REML, but only because this is the default in lme and I did
not
>    change it.
>
>    As I have not applied a lack-of-fit test before, I would really be glad
for
>    any help!
>
>    Best regards,
>    John
> ______________________________________________
> R-help at 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.