[R-sig-ME] (no subject)

Discussion:

Peter Paprzycki

2018-08-02 03:30:15 UTC

This is very basic, is there a way to specify in lmer function that I would
like to run my grouping variable as a fixed factor only, without reverting
to lm or plm functions. If one does not specify a random variable, one gets
the error message with lmer function; something that is equivalent to the
statement, "index = "grouping variable", model = "within"" with the plm
function.

<http://www.avg.com/email-signature?utm_medium=email&utm_source=link&utm_campaign=sig-email&utm_content=webmail>
Virus-free.
www.avg.com
<http://www.avg.com/email-signature?utm_medium=email&utm_source=link&utm_campaign=sig-email&utm_content=webmail>
<#DAB4FAD8-2DD7-40BB-A1B8-4E2AA1F9FDF2>

[[alternative HTML version deleted]]

Ben Bolker

2018-08-02 03:57:38 UTC

Permalink

(please keep r-sig-mixed-models in the Cc:)

I'm pretty sure that lmer and lm models are commensurate, in case that
helps. Here's an example rigged to make the random-effects variance
equal to zero, so we can check that the log-likelihoods etc. are identical.

set.seed(101)
dd <- data.frame(y=rnorm(20),x=rnorm(20),f=factor(rep(1:2,10)))
library(lme4)
m1 <- lmer(y~x+(1|f),data=dd,REML=FALSE) ## estimated sigma^2_f=0
m2 <- lm(y~x,data=dd)
all.equal(c(logLik(m1)),c(logLik(m2))) ## TRUE
all.equal(fixef(m1),coef(m2))
anova(m1,m2)

Thank you. Oh, was just trying to compare my random-effects model to the
one where my grouping variable (schools) is treated as fixed.
Peter
<http://www.avg.com/email-signature?utm_medium=email&utm_source=link&utm_campaign=sig-email&utm_content=webmail>
Virus-free. www.avg.com
<http://www.avg.com/email-signature?utm_medium=email&utm_source=link&utm_campaign=sig-email&utm_content=webmail>
<#DAB4FAD8-2DD7-40BB-A1B8-4E2AA1F9FDF2>
I'm not 100% sure I understand the question, but I think the answer is
"no": lmer cannot fit a model that doesn't contain any random effects.
Perhaps you can give more context as to why it won't work for you to
revert to lm() or plm() in these cases?

Post by Peter Paprzycki
This is very basic, is there a way to specify in lmer function

that I would

Post by Peter Paprzycki
like to run my grouping variable as a fixed factor only, without

reverting

Post by Peter Paprzycki
to lm or plm functions. If one does not specify a random variable,

one gets

Post by Peter Paprzycki
the error message with lmer function; something that is equivalent

to the

Post by Peter Paprzycki
statement, "index = "grouping variable", model = "within"" with

the plm

Post by Peter Paprzycki
function.

<http://www.avg.com/email-signature?utm_medium=email&utm_source=link&utm_campaign=sig-email&utm_content=webmail
<http://www.avg.com/email-signature?utm_medium=email&utm_source=link&utm_campaign=sig-email&utm_content=webmail>>

Post by Peter Paprzycki
Virus-free.
www.avg.com <http://www.avg.com>

Post by Peter Paprzycki
<#DAB4FAD8-2DD7-40BB-A1B8-4E2AA1F9FDF2>
[[alternative HTML version deleted]]
_______________________________________________
https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models

<https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models>
--
Peter Paprzycki, Ph.D.
Visiting Assistant Professor
Research Support Center Manager
Educational Research and Administration
College of Education and Psychology
The University of Southern Mississippi
USM Box 5093; 118 College Drive
Hattiesburg, Mississippi 39406-0001
tel. (601)-266-4708
Sidere mens eadum mutato

Peter Paprzycki

2018-08-02 04:08:10 UTC

Permalink

Perfect. Thank you. It is good to know that we can specify the
random-effects variance
equal to zero. Thanks.
Peter

<http://www.avg.com/email-signature?utm_medium=email&utm_source=link&utm_campaign=sig-email&utm_content=webmail>
Virus-free.
www.avg.com
<http://www.avg.com/email-signature?utm_medium=email&utm_source=link&utm_campaign=sig-email&utm_content=webmail>
<#DAB4FAD8-2DD7-40BB-A1B8-4E2AA1F9FDF2>

Post by Ben Bolker
(please keep r-sig-mixed-models in the Cc:)
I'm pretty sure that lmer and lm models are commensurate, in case that
helps. Here's an example rigged to make the random-effects variance
equal to zero, so we can check that the log-likelihoods etc. are identical.
set.seed(101)
dd <- data.frame(y=rnorm(20),x=rnorm(20),f=factor(rep(1:2,10)))
library(lme4)
m1 <- lmer(y~x+(1|f),data=dd,REML=FALSE) ## estimated sigma^2_f=0
m2 <- lm(y~x,data=dd)
all.equal(c(logLik(m1)),c(logLik(m2))) ## TRUE
all.equal(fixef(m1),coef(m2))
anova(m1,m2)

Thank you. Oh, was just trying to compare my random-effects model to the
one where my grouping variable (schools) is treated as fixed.
Peter
<http://www.avg.com/email-signature?utm_medium=email&

utm_source=link&utm_campaign=sig-email&utm_content=webmail>

Virus-free. www.avg.com
<http://www.avg.com/email-signature?utm_medium=email&

utm_source=link&utm_campaign=sig-email&utm_content=webmail>

<#DAB4FAD8-2DD7-40BB-A1B8-4E2AA1F9FDF2>
I'm not 100% sure I understand the question, but I think the

answer is

"no": lmer cannot fit a model that doesn't contain any random

effects.

Perhaps you can give more context as to why it won't work for you to
revert to lm() or plm() in these cases?

Post by Peter Paprzycki
This is very basic, is there a way to specify in lmer function

that I would

Post by Peter Paprzycki
like to run my grouping variable as a fixed factor only, without

reverting

Post by Peter Paprzycki
to lm or plm functions. If one does not specify a random variable,

one gets

Post by Peter Paprzycki
the error message with lmer function; something that is equivalent

to the

Post by Peter Paprzycki
statement, "index = "grouping variable", model = "within"" with

the plm

Post by Peter Paprzycki
function.

<http://www.avg.com/email-signature?utm_medium=email&

utm_source=link&utm_campaign=sig-email&utm_content=webmail

<http://www.avg.com/email-signature?utm_medium=email&

utm_source=link&utm_campaign=sig-email&utm_content=webmail>>

Post by Peter Paprzycki
Virus-free.
www.avg.com <http://www.avg.com>

<http://www.avg.com/email-signature?utm_medium=email&

utm_source=link&utm_campaign=sig-email&utm_content=webmail

<http://www.avg.com/email-signature?utm_medium=email&

utm_source=link&utm_campaign=sig-email&utm_content=webmail>>

--
Peter Paprzycki, Ph.D.
Visiting Assistant Professor
Research Support Center Manager

Educational Research and Administration
College of Education and Psychology
The University of Southern Mississippi
USM Box 5093; 118 College Drive
Hattiesburg, Mississippi 39406-0001
tel. (601)-266-4708
email: ***@usm.edu

Sidere mens eadum mutato

[[alternative HTML version deleted]]

Peter Paprzycki

2018-08-02 04:24:10 UTC

Permalink

Sorry, you estimated it to be very close to zero, I see.
Peter

Post by Peter Paprzycki
Perfect. Thank you. It is good to know that we can specify the
random-effects variance
equal to zero. Thanks.
Peter
<http://www.avg.com/email-signature?utm_medium=email&utm_source=link&utm_campaign=sig-email&utm_content=webmail> Virus-free.
www.avg.com
<http://www.avg.com/email-signature?utm_medium=email&utm_source=link&utm_campaign=sig-email&utm_content=webmail>
<#m_981606773260833816_DAB4FAD8-2DD7-40BB-A1B8-4E2AA1F9FDF2>

Thank you. Oh, was just trying to compare my random-effects model to the
one where my grouping variable (schools) is treated as fixed.
Peter
<http://www.avg.com/email-signature?utm_medium=email&utm_

source=link&utm_campaign=sig-email&utm_content=webmail>

Virus-free. www.avg.com
<http://www.avg.com/email-signature?utm_medium=email&utm_

source=link&utm_campaign=sig-email&utm_content=webmail>

<#DAB4FAD8-2DD7-40BB-A1B8-4E2AA1F9FDF2>
I'm not 100% sure I understand the question, but I think the

answer is

"no": lmer cannot fit a model that doesn't contain any random

effects.

Perhaps you can give more context as to why it won't work for you to
revert to lm() or plm() in these cases?

Post by Peter Paprzycki
This is very basic, is there a way to specify in lmer function

that I would

Post by Peter Paprzycki
like to run my grouping variable as a fixed factor only, without

reverting

Post by Peter Paprzycki
to lm or plm functions. If one does not specify a random variable,

one gets

Post by Peter Paprzycki
the error message with lmer function; something that is equivalent

to the

Post by Peter Paprzycki
statement, "index = "grouping variable", model = "within"" with

the plm

Post by Peter Paprzycki
function.

<http://www.avg.com/email-signature?utm_medium=email&utm_

source=link&utm_campaign=sig-email&utm_content=webmail

<http://www.avg.com/email-signature?utm_medium=email&utm_

source=link&utm_campaign=sig-email&utm_content=webmail>>

Post by Peter Paprzycki
Virus-free.
www.avg.com <http://www.avg.com>

<http://www.avg.com/email-signature?utm_medium=email&utm_

source=link&utm_campaign=sig-email&utm_content=webmail

<http://www.avg.com/email-signature?utm_medium=email&utm_

source=link&utm_campaign=sig-email&utm_content=webmail>>

--
Peter Paprzycki, Ph.D.
Visiting Assistant Professor
Research Support Center Manager
Educational Research and Administration
College of Education and Psychology
The University of Southern Mississippi
USM Box 5093; 118 College Drive
Hattiesburg, Mississippi 39406-0001
tel. (601)-266-4708
Sidere mens eadum mutato

Ben Bolker

2018-08-02 05:26:27 UTC

Permalink

Yes. I think you can specify a fixed residual variance in blme::blmer, but
not to exactly zero.

Post by Peter Paprzycki
Sorry, you estimated it to be very close to zero, I see.
Peter
On Wed, Aug 1, 2018 at 11:08 PM, Peter Paprzycki <

Post by Peter Paprzycki
Perfect. Thank you. It is good to know that we can specify the
random-effects variance
equal to zero. Thanks.
Peter
<http://www.avg.com/email-signature?utm_medium=email&utm_source=link&utm_campaign=sig-email&utm_content=webmail> Virus-free.
www.avg.com
<http://www.avg.com/email-signature?utm_medium=email&utm_source=link&utm_campaign=sig-email&utm_content=webmail>
<#m_3076315220315282584_m_981606773260833816_DAB4FAD8-2DD7-40BB-A1B8-4E2AA1F9FDF2>

Thank you. Oh, was just trying to compare my random-effects model to

the

one where my grouping variable (schools) is treated as fixed.
Peter
<

http://www.avg.com/email-signature?utm_medium=email&utm_source=link&utm_campaign=sig-email&utm_content=webmail

Virus-free. www.avg.com
<

http://www.avg.com/email-signature?utm_medium=email&utm_source=link&utm_campaign=sig-email&utm_content=webmail

<#DAB4FAD8-2DD7-40BB-A1B8-4E2AA1F9FDF2>
I'm not 100% sure I understand the question, but I think the

answer is

"no": lmer cannot fit a model that doesn't contain any random

effects.

Perhaps you can give more context as to why it won't work for you

revert to lm() or plm() in these cases?

Post by Peter Paprzycki
This is very basic, is there a way to specify in lmer function

that I would

Post by Peter Paprzycki
like to run my grouping variable as a fixed factor only, without

reverting

Post by Peter Paprzycki
to lm or plm functions. If one does not specify a random

variable,

one gets

Post by Peter Paprzycki
the error message with lmer function; something that is

equivalent

to the

Post by Peter Paprzycki
statement, "index = "grouping variable", model = "within"" with

the plm

Post by Peter Paprzycki
function.

http://www.avg.com/email-signature?utm_medium=email&utm_source=link&utm_campaign=sig-email&utm_content=webmail

Post by Peter Paprzycki
Virus-free.
www.avg.com <http://www.avg.com>

http://www.avg.com/email-signature?utm_medium=email&utm_source=link&utm_campaign=sig-email&utm_content=webmail

--
Peter Paprzycki, Ph.D.
Visiting Assistant Professor
Research Support Center Manager
Educational Research and Administration
College of Education and Psychology
The University of Southern Mississippi
USM Box 5093; 118 College Drive
Hattiesburg, Mississippi 39406-0001
tel. (601)-266-4708
Sidere mens eadum mutato

[[alternative HTML version deleted]]

Martin Maechler

2018-08-02 06:53:59 UTC

Permalink

Post by Ben Bolker

Post by Peter Paprzycki

Post by Ben Bolker
I'm pretty sure that lmer and lm models are commensurate, in case that
helps. Here's an example rigged to make the random-effects variance
equal to zero, so we can check that the log-likelihoods etc. are identical.
set.seed(101)
dd <- data.frame(y=rnorm(20),x=rnorm(20),f=factor(rep(1:2,10)))
library(lme4)
m1 <- lmer(y~x+(1|f),data=dd,REML=FALSE) ## estimated sigma^2_f=0
m2 <- lm(y~x,data=dd)
all.equal(c(logLik(m1)),c(logLik(m2))) ## TRUE
all.equal(fixef(m1),coef(m2))
anova(m1,m2)

Then, Peter replied

Post by Ben Bolker

Post by Peter Paprzycki
Sorry, you estimated it to be very close to zero, I see.
Peter

Yes. I think you can specify a fixed residual variance in blme::blmer, but
not to exactly zero.

Peter: it is estimated to be *exactly* zero, not just close to

Post by Ben Bolker
VarCorr(m1)$f == 0

(Intercept)
(Intercept) TRUE
(yes, these are always matrices, here of dimension 1x1)

This has been one of the major features of lme4::lmer() wrt to nlme::lme()
that \hat{\sigma_j^2} = 0 is naturally possible
because of the parametrization used.

Martin

Peter Paprzycki

2018-08-02 21:47:08 UTC

Permalink

Thank you, Martin.
Peter

<http://www.avg.com/email-signature?utm_medium=email&utm_source=link&utm_campaign=sig-email&utm_content=webmail>
Virus-free.
www.avg.com
<http://www.avg.com/email-signature?utm_medium=email&utm_source=link&utm_campaign=sig-email&utm_content=webmail>
<#DAB4FAD8-2DD7-40BB-A1B8-4E2AA1F9FDF2>

Post by Martin Maechler

Post by Ben Bolker

Post by Peter Paprzycki

Then, Peter replied

Post by Ben Bolker

Post by Peter Paprzycki
Sorry, you estimated it to be very close to zero, I see.
Peter

Yes. I think you can specify a fixed residual variance in

blme::blmer, but

Post by Ben Bolker
not to exactly zero.

Peter: it is estimated to be *exactly* zero, not just close to

Post by Ben Bolker
VarCorr(m1)$f == 0

(Intercept)
(Intercept) TRUE
(yes, these are always matrices, here of dimension 1x1)
This has been one of the major features of lme4::lmer() wrt to nlme::lme()
that \hat{\sigma_j^2} = 0 is naturally possible
because of the parametrization used.
Martin