[R-sig-ME] R-sig-mixed-models Digest, Vol 27, Issue 36
Jarrod Hadfield
j.hadfield at ed.ac.uk
Fri Apr 3 14:55:22 CEST 2009
Hi,
There's a reasonable amount of data (the least replicated category has
24 data points - Year2Egg3) so I would suggest this is not so much to
do with over-fitting but more to do with the underlying probabilities
for some categories being extreme. The paper that Ken sent round
provides a Bayesian solution to the problem by placing weak priors on
the fixed effects. Cauchy/Scaled-t priors were recommended in that
paper, and this can presumably be done in WinBUGS and maybe JAGS(?).
Normal priors can be used in MCMCglmm or rhierBinLogit in the bayesm
package, and these can be set to be weakly informative on the
probability scale. For MCMCglmm at least, the choice of prior will
depend on your choice of residual variance.
Cheers,
Jarrod
On 3 Apr 2009, at 13:17, Luciano La Sala wrote:
>
> Hello Jarrod,
>
> Thanks for the tip. Here is my tabulated data so that maybe someone
> can tell me wether this is numerical instability or over-fitting
> that I'm seing in my model.
>
> , , Year = 1
>
> Hatching_Order
> Hatching 1 2 3
> 0 45 33 25
> 1 2 0 5
>
> , , Year = 2
>
> Hatching_Order
> Hatching 1 2 3
> 0 64 34 13
> 1 3 12 11
>
> From this table, I was not able to diagnose mu model's problem, but
> hopefully someone will!
>
> Best
>
> Luciano
>
>
>
>
> --- El jue 2-abr-09, Jarrod Hadfield <j.hadfield at ed.ac.uk> escribió:
>
> De: Jarrod Hadfield <j.hadfield at ed.ac.uk>
> Asunto: Re: R-sig-mixed-models Digest, Vol 27, Issue 36
> Para: lucianolasala at yahoo.com.ar
> Cc: r-sig-mixed-models at r-project.org, ken at kjbeath.com.au
> Fecha: jueves, 2 de abril de 2009, 10:36 am
>
> Hi Luciano,
>
> Perhaps you could tabulate your results so we could see whether this
> really is an over-fitting problem or a numerical problem. Something
> like:
>
> table(lHatching, HatchOrder, Year)
>
> should do it.
>
> Cheers,
>
> Jarrod
>
>
> On 1 Apr 2009, at 22:04, Luciano La Sala wrote:
>
>>
>> Dear Ken and Jarrod,
>>
>> Thank you very much for shedding some light on my problem!
>> Besides the points you've made, I was told that I could avoid over-
>> fitting in my second model by treating the interaction term as two
>> continuous variables instead of categorical ones
>> (HatchingOrder*Year). Well, after doind so, the output says "Error
>> in asMethod(object) : matrix is not symmetric [1,2]", so I guess
>> that that is not a feasible solution to the problem either, and I
>> wonder how far should one go to find one?
>>
>> If my model fell prey to Hauck-Donner effect - which may be causing
>> the std.. errors to be overestimated and the signifficance of the
>> effect to be missed - maybe I just should stop turturing my data and
>> admit that it is sparse?
>>
>> Should you come up with any ideas, I'd be glad to hear them.
>>
>> Luciano
>>
>>
>>
>>
>> --- El lun 30-mar-09, r-sig-mixed-models-request at r-project.org
> <r-sig-mixed-models-request at r-project.org
>>> escribió:
>>
>>> De: r-sig-mixed-models-request at r-project.org
> <r-sig-mixed-models-request at r-project.org
>>>>
>>> Asunto: R-sig-mixed-models Digest, Vol 27, Issue 36
>>> Para: r-sig-mixed-models at r-project.org
>>> Fecha: lunes, 30 de marzo de 2009, 6:27 pm
>>> Send R-sig-mixed-models mailing list submissions to
>>> r-sig-mixed-models at r-project.org
>>>
>>> To subscribe or unsubscribe via the World Wide Web, visit
>>> https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models
>>> or, via email, send a message with subject or body
>>> 'help' to
>>> r-sig-mixed-models-request at r-project.org
>>>
>>> You can reach the person managing the list at
>>> r-sig-mixed-models-owner at r-project.org
>>>
>>> When replying, please edit your Subject line so it is more
>>> specific
>>> than "Re: Contents of R-sig-mixed-models
>>> digest..."
>>>
>>>
>>> Today's Topics:
>>>
>>> 1. Re: Can interaction term cause Estimates and Std.
>>> Errors to
>>> be too large? (Ken Beath)
>>> 2. Re: Can interaction term cause Estimates and Std.
>>> Errors to
>>> be too large? (Jarrod Hadfield)
>>> 3. Re: Can interaction term cause Estimates and Std.
>>> Errors to
>>> be too large? (Ken Beath)
>>> 4. Meta-analysis using lmer (Yu-Kang Tu)
>>> 5. Mixed Model for Travel Distance (Chuck Cleland)
>>> 6. Re: Mixed Model for Travel Distance (Dimitris
>>> Rizopoulos)
>>>
>>>
>>> ----------------------------------------------------------------------
>>>
>>> Message: 1
>>> Date: Mon, 30 Mar 2009 21:08:58 +1100
>>> From: Ken Beath <ken at kjbeath.com.au>
>>> Subject: Re: [R-sig-ME] Can interaction term cause
>>> Estimates and Std.
>>> Errors to be too large?
>>> To: Jarrod Hadfield <j.hadfield at ed.ac.uk>
>>> Cc: r-sig-mixed-models at r-project.org,
>>> lucianolasala at yahoo.com.ar
>>> Message-ID:
>>> <68347D8C-AE1C-48BF-B194-D7E1DDD50B50 at kjbeath.com.au>
>>> Content-Type: text/plain; charset=US-ASCII; format=flowed;
>>> delsp=yes
>>>
>>> I meant overfitting in the sense of trying to fit too
>>> complex a model,
>>> which is the same as what you are describing. Gelman has
>>> some papers
>>> on the use of priors, one is
>>>
> http://projecteuclid.org/DPubS?service=UI&version=1.0&verb=Display&handle=euclid.aoas/1231424214
>>>
>>> In the case of complete separation the results seem to be
>>> very
>>> dependent on the prior which doesn't look to be a good
>>> thing. It would
>>> appear much better to admit that there is insufficient data
>>> to perform
>>> the analysis.
>>>
>>> Ken
>>>
>>>
>>> On 30/03/2009, at 7:48 PM, Jarrod Hadfield wrote:
>>>
>>>> Hi,
>>>>
>>>> I think it unlikely that the problem arises through
>>> overfitting in
>>>> the sense that there are too many parameters for the
>>> amount of
>>>> data. It's more likely that the underlying
>>> probabilities really are
>>>> extreme for some categories causing what are also
>>> known as "extreme
>>>> category problems" (eg Miztal 1998 J. Dairy
>>> Science 72 1557-1568):
>>>> the binary variable in one or more groups is always 0
>>> or 1, even
>>>> though there are probably many eggs in most
>>> categories. A solution
>>>> to this type of problem is to place an informative
>>> prior on the
>>>> fixed effects to stop them wandering into extreme
>>> values on the
>>>> logit scale. For the purist this may be anathema, but
>>> as a practical
>>>> solution it seems to work quite well. Having a normal
>>> prior on the
>>>> logit scale with mean zero and variance pi, is the
>>> closest (I
>>>> think?) to a uniform prior on the probability scale.
>>> If there are
>>>> more elegant solutions to the problem I'd be
>>> interested to hear
>>>> about them.
>>>>
>>>> Cheers,
>>>>
>>>> Jarrod
>>>>
>>>>
>>>>
>>>> --
>>>> The University of Edinburgh is a charitable body,
>>> registered in
>>>> Scotland, with registration number SC005336.
>>>>
>>>>
>>>
>>>
>>>
>>> ------------------------------
>>>
>>> Message: 2
>>> Date: Mon, 30 Mar 2009 11:21:39 +0100
>>> From: Jarrod Hadfield <j.hadfield at ed.ac.uk>
>>> Subject: Re: [R-sig-ME] Can interaction term cause
>>> Estimates and Std.
>>> Errors to be too large?
>>> To: Ken Beath <ken at kjbeath.com.au>
>>> Cc: r-sig-mixed-models at r-project.org,
>>> lucianolasala at yahoo.com.ar
>>> Message-ID:
>>> <AC5DBE6D-DEEF-47A2-90BE-071FE9CBCC79 at ed.ac.uk>
>>> Content-Type: text/plain; charset="us-ascii";
>>> Format="flowed";
>>> DelSp="yes"
>>>
>>> Hi Ken,
>>>
>>> Thanks for the reference, it looks interesting. I disagree
>>> that
>>> Luciano's second model should be classified as over
>>> fitting. Imagine
>>> this....
>>>
>>> y<-rbinom(100, 1, c(0.001, 0.999))
>>> x<-gl(2,1,100)
>>>
>>> summary(glm(y~1, family="binomial"))
>>> summary(glm(y~x, family="binomial"))
>>>
>>> There is a very high probability of complete separation,
>>> the second
>>> model gives non-significant p-values for the effect of x,
>>> but I think
>>> it would be a mistake to say the 2nd model is over-fitted
>>> and should
>>> be avoided.
>>>
>>> Cheers,
>>>
>>> Jarrod
>>>
>>>
>>> On 30 Mar 2009, at 11:08, Ken Beath wrote:
>>>
>>>> I meant overfitting in the sense of trying to fit too
>>> complex a
>>>> model, which is the same as what you are describing.
>>> Gelman has some
>>>> papers on the use of priors, one is
>>>
> http://projecteuclid.org/DPubS?service=UI&version=1.0&verb=Display&handle=euclid.aoas/1231424214
>>>
>>>> In the case of complete separation the results seem
>>> to be very
>>>> dependent on the prior which doesn't look to be a
>>> good thing. It
>>>> would appear much better to admit that there is
>>> insufficient data to
>>>> perform the analysis.
>>>>
>>>> Ken
>>>>
>>>>
>>>> On 30/03/2009, at 7:48 PM, Jarrod Hadfield wrote:
>>>>
>>>>> Hi,
>>>>>
>>>>> I think it unlikely that the problem arises
>>> through overfitting in
>>>>> the sense that there are too many parameters for
>>> the amount of
>>>>> data. It's more likely that the underlying
>>> probabilities really
>>>>> are extreme for some categories causing what are
>>> also known as
>>>>> "extreme category problems" (eg Miztal
>>> 1998 J. Dairy Science 72
>>>>> 1557-1568): the binary variable in one or more
>>> groups is always 0
>>>>> or 1, even though there are probably many eggs in
>>> most
>>>>> categories. A solution to this type of problem is
>>> to place an
>>>>> informative prior on the fixed effects to stop
>>> them wandering into
>>>>> extreme values on the logit scale. For the purist
>>> this may be
>>>>> anathema, but as a practical solution it seems to
>>> work quite well.
>>>>> Having a normal prior on the logit scale with mean
>>> zero and
>>>>> variance pi, is the closest (I think?) to a
>>> uniform prior on the
>>>>> probability scale. If there are more elegant
>>> solutions to the
>>>>> problem I'd be interested to hear about them.
>>>>>
>>>>> Cheers,
>>>>>
>>>>> Jarrod
>>>>>
>>>>>
>>>>>
>>>>> --
>>>>> The University of Edinburgh is a charitable body,
>>> registered in
>>>>> Scotland, with registration number SC005336.
>>>>>
>>>>>
>>>>
>>>>
>>>
>>> -------------- next part --------------
>>> An embedded and charset-unspecified text was scrubbed...
>>> Name: not available
>>> URL:
>>>
> <https://stat.ethz.ch/pipermail/r-sig-mixed-models/attachments/20090330/df46ef59/attachment-0001.pl
>
>>>>
>>>
>>> ------------------------------
>>>
>>> Message: 3
>>> Date: Mon, 30 Mar 2009 21:50:53 +1100
>>> From: Ken Beath <ken at kjbeath.com.au>
>>> Subject: Re: [R-sig-ME] Can interaction term cause
>>> Estimates and Std.
>>> Errors to be too large?
>>> To: Jarrod Hadfield <j.hadfield at ed.ac.uk>
>>> Cc: "r-sig-mixed-models at r-project.org Models"
>>> <r-sig-mixed-models at r-project.org>,
>>> lucianolasala at yahoo.com.ar
>>> Message-ID:
>>> <AF2C8F6D-8857-4855-BAED-8DD9B525212D at kjbeath.com.au>
>>> Content-Type: text/plain; charset=US-ASCII; format=flowed;
>>> delsp=yes
>>>
>>> On 30/03/2009, at 9:21 PM, Jarrod Hadfield wrote:
>>>
>>>> Hi Ken,
>>>>
>>>> Thanks for the reference, it looks interesting. I
>>> disagree that
>>>> Luciano's second model should be classified as
>>> over fitting. Imagine
>>>> this....
>>>>
>>>> y<-rbinom(100, 1, c(0.001, 0.999))
>>>> x<-gl(2,1,100)
>>>>
>>>> summary(glm(y~1, family="binomial"))
>>>> summary(glm(y~x, family="binomial"))
>>>>
>>>> There is a very high probability of complete
>>> separation, the second
>>>> model gives non-significant p-values for the effect of
>>> x, but I
>>>> think it would be a mistake to say the 2nd model is
>>> over-fitted and
>>>> should be avoided.
>>>>
>>>> Cheers,
>>>>
>>>> Jarrod
>>>>
>>>
>>> My original posting said "usually" and obviously
>>> you can create data
>>> with perfect or almost perfect correlation over a large
>>> table, but in
>>> practice it commonly happens because there is a small
>>> table. One good
>>> reason for producing some descriptive tables before
>>> fitting.
>>>
>>> Ken
>>>
>>>
>>>>
>>>> On 30 Mar 2009, at 11:08, Ken Beath wrote:
>>>>
>>>>> I meant overfitting in the sense of trying to fit
>>> too complex a
>>>>> model, which is the same as what you are
>>> describing. Gelman has
>>>>> some papers on the use of priors, one is
>>>
> http://projecteuclid.org/DPubS?service=UI&version=1.0&verb=Display&handle=euclid.aoas/1231424214
>>>
>>>>> In the case of complete separation the results
>>> seem to be very
>>>>> dependent on the prior which doesn't look to
>>> be a good thing. It
>>>>> would appear much better to admit that there is
>>> insufficient data
>>>>> to perform the analysis.
>>>>>
>>>>> Ken
>>>>>
>>>>>
>>>>> On 30/03/2009, at 7:48 PM, Jarrod Hadfield wrote:
>>>>>
>>>>>> Hi,
>>>>>>
>>>>>> I think it unlikely that the problem arises
>>> through overfitting in
>>>>>> the sense that there are too many parameters
>>> for the amount of
>>>>>> data. It's more likely that the
>>> underlying probabilities really
>>>>>> are extreme for some categories causing what
>>> are also known as
>>>>>> "extreme category problems" (eg
>>> Miztal 1998 J. Dairy Science 72
>>>>>> 1557-1568): the binary variable in one or more
>>> groups is always 0
>>>>>> or 1, even though there are probably many eggs
>>> in most
>>>>>> categories. A solution to this type of
>>> problem is to place an
>>>>>> informative prior on the fixed effects to stop
>>> them wandering into
>>>>>> extreme values on the logit scale. For the
>>> purist this may be
>>>>>> anathema, but as a practical solution it seems
>>> to work quite
>>>>>> well. Having a normal prior on the logit
>>> scale with mean zero and
>>>>>> variance pi, is the closest (I think?) to a
>>> uniform prior on the
>>>>>> probability scale. If there are more elegant
>>> solutions to the
>>>>>> problem I'd be interested to hear about
>>> them.
>>>>>>
>>>>>> Cheers,
>>>>>>
>>>>>> Jarrod
>>>>>>
>>>>>>
>>>>>>
>>>>>> --
>>>>>> The University of Edinburgh is a charitable
>>> body, registered in
>>>>>> Scotland, with registration number SC005336.
>>>>>>
>>>>>>
>>>>>
>>>>>
>>>>
>>>> The University of Edinburgh is a charitable body,
>>> registered in
>>>> Scotland, with registration number SC005336.
>>>
>>>
>>>
>>> ------------------------------
>>>
>>> Message: 4
>>> Date: Mon, 30 Mar 2009 18:16:46 +0100
>>> From: Yu-Kang Tu <Y.K.Tu at leeds.ac.uk>
>>> Subject: [R-sig-ME] Meta-analysis using lmer
>>> To: "'r-sig-mixed-models at r-project.org'"
>>> <r-sig-mixed-models at r-project.org>
>>> Message-ID:
>>>
> <7131EF1EF27893479833FCF59E7AAC440B3E818BE1 at HERMES9.ds.leeds.ac.uk>
>>> Content-Type: text/plain; charset="us-ascii"
>>>
>>>
>>> Hi,
>>>
>>> I am trying to use lme and lmer to do random effects
>>> meta-analysis as described in Hox (2002) and UCLA website:
>>> http://www.ats.ucla.edu/stat/mlwin/examples/ma_hox/chapter8.htm
>>>
>>> Basically, what I want to do is to constraint one residual
>>> error variance to be unity and use the inverse of standard
>>> errors as the covariate (weight) for this variance. And an
>>> additional random effects terms is used to estimate the
>>> between-study variation. I did take a look at the Pinheiro
>>> & Bates book on varFunc, but unfortunately, I cannot
>>> figure out how this can be done. Any suggestions/advices
>>> will be greatly appreciated. Many thanks.
>>>
>>> Yu-Kang
>>> --------------------------------------------
>>> Dr Yu-Kang Tu
>>> Senior Clinical Research Fellow
>>> Division of Biostatistics, Centre for Epidemiology and
>>> Biostatistics
>>> Leeds Institute of Genetics, Health and Therapeutics, and
>>> Department of Periodontology, Leeds Dental Institute
>>> Room 8.01, Level 8, Worsley Building,
>>> Clarendon Way
>>> University of Leeds, LS2 9JT
>>> Email: y.k.tu at leeds.ac.uk
>>> Tel: +44 (0) 113 3431877
>>> Fax: +44 (0) 113 3434877
>>>
>>>
>>>
>>> ------------------------------
>>>
>>> Message: 5
>>> Date: Mon, 30 Mar 2009 13:29:15 -0400
>>> From: Chuck Cleland <ccleland at optonline.net>
>>> Subject: [R-sig-ME] Mixed Model for Travel Distance
>>> To: r-sig-mixed-models at r-project.org
>>> Message-ID: <49D1016B.8050808 at optonline.net>
>>> Content-Type: text/plain; charset=ISO-8859-1
>>>
>>> Hello:
>>> I am attempting to model the distance that clients travel
>>> to a
>>> treatment program. There are 14385 clients nested in 83
>>> treatment
>>> programs (the grouping factor). The raw data are in miles
>>> driven from
>>> the client's residence to the treatment program. A
>>> natural logarithm
>>> transformation of miles driven works well to reduce the
>>> positive skew in
>>> miles driven. I fit a model with lme() that looks like
>>> this:
>>>
>>> Linear mixed-effects model fit by REML
>>> Data: dist.df
>>> AIC BIC logLik
>>> 38145.37 38319.54 -19049.68
>>>
>>> Random effects:
>>> Formula: ~1 | PROGRAM
>>> (Intercept) Residual
>>> StdDev: 0.4268969 0.8988483
>>>
>>> Fixed effects: log(DIST5DZ + 1) ~ QUAD + BEAL_TRI +
>>> log(RZIPAREA + 1) +
>>> log(PZIPAREA + 1) + AGE.TRI + P3GEND + P5RACEX + EMPLD +
>>> P13REASN +
>>> METHFST + URGE.DI + WDRAW.DI + RX_30 + P7HR30
>>>
>>> Value Std.Error DF
>>> t-value p-value
>>> (Intercept) 1.2235603 0.13153231 14288
>>> 9.30236 0.0000
>>> QUADSouthEast 0.2100666 0.13200891 76
>>> 1.59131 0.1157
>>> QUADMidWest 0.2760390 0.15709516 76
>>> 1.75715 0.0829
>>> QUADWest -0.1655914 0.15536003 76
>>> -1.06586 0.2899
>>> BEAL_TRI250K-1M -0.0264939 0.11724713 76
>>> -0.22597 0.8218
>>> BEAL_TRI<250K -0.0965256 0.16399464 76
>>> -0.58859 0.5579
>>> log(RZIPAREA + 1) 0.2965304 0.00757138 14288
>>> 39.16463 0.0000
>>> log(PZIPAREA + 1) -0.0042061 0.04413826 76
>>> -0.09529 0.9243
>>> AGE.TRI30-43 -0.0309444 0.01789442 14288
>>> -1.72927 0.0838
>>> AGE.TRI43-83 -0.1281177 0.02168648 14288
>>> -5.90772 0.0000
>>> P3GENDFemale -0.0195289 0.01632703 14288
>>> -1.19611 0.2317
>>> P5RACEXLatino -0.3527584 0.02904416 14288
>>> -12.14559 0.0000
>>> P5RACEXBlack -0.5485861 0.03306146 14288
>>> -16.59292 0.0000
>>> P5RACEXOther -0.1580669 0.04811350 14288
>>> -3.28529 0.0010
>>> EMPLDYes 0.0098856 0.01650635 14288
>>> 0.59890 0.5493
>>> P13REASNYes -0.0095057 0.01650128 14288
>>> -0.57606 0.5646
>>> METHFSTYes 0.0073478 0.01672948 14288
>>> 0.43921 0.6605
>>> URGE.DI Strong-VeryStrong -0.0272769 0.02400068 14288
>>> -1.13651 0.2558
>>> WDRAW.DISevere-VerySevere 0.0012054 0.01810067 14288
>>> 0.06659 0.9469
>>> RX_30Yes 0.0934411 0.02165389 14288
>>> 4.31521 0.0000
>>> P7HR30Yes -0.0408189 0.02256842 14288
>>> -1.80867 0.0705
>>>
>>> Standardized Within-Group Residuals:
>>> Min Q1 Med Q3 Max
>>> -4.17921640 -0.49875991 0.08672984 0.59020542 4.54644432
>>>
>>> Number of Observations: 14385
>>> Number of Groups: 83
>>>
>>> I would like to summarize the fixed effects in terms of
>>> miles rather
>>> than log(miles + 1). How can that be done? Are there
>>> common
>>> generalized linear mixed models for miles driven that would
>>> avoid the
>>> transformation and allow effects to be presented in miles?
>>>
>>> thanks,
>>>
>>> Chuck
>>>
>>> --
>>> Chuck Cleland, Ph.D.
>>> NDRI, Inc. (www.ndri.org)
>>> 71 West 23rd Street, 8th floor
>>> New York, NY 10010
>>> tel: (212) 845-4495 (Tu, Th)
>>> tel: (732) 512-0171 (M, W, F)
>>> fax: (917) 438-0894
>>>
>>>
>>>
>>> ------------------------------
>>>
>>> Message: 6
>>> Date: Mon, 30 Mar 2009 20:26:38 +0200
>>> From: Dimitris Rizopoulos <d.rizopoulos at erasmusmc.nl>
>>> Subject: Re: [R-sig-ME] Mixed Model for Travel Distance
>>> To: Chuck Cleland <ccleland at optonline.net>
>>> Cc: r-sig-mixed-models at r-project.org
>>> Message-ID: <49D10EDE.8050501 at erasmusmc.nl>
>>> Content-Type: text/plain; charset=ISO-8859-1; format=flowed
>>>
>>> well, if you're only interested in the fixed effects,
>>> then you can also
>>> use a Generalized Estimating Equations approach that does
>>> not make a
>>> parametric assumption for the distribution of your error
>>> terms, e.g.,
>>> have a look at the 'geepack' package. Furthermore
>>> and in case it is
>>> relevant for your application, in GEE the estimated
>>> parameters will have
>>> a population interpretation, contrary to the GLMMs approach
>>> in which
>>> they will have a conditional on the random effects
>>> interpretation.
>>>
>>>
>>> I hope it helps.
>>>
>>> Best,
>>> Dimitris
>>>
>>>
>>> Chuck Cleland wrote:
>>>> Hello:
>>>> I am attempting to model the distance that clients
>>> travel to a
>>>> treatment program. There are 14385 clients nested in
>>> 83 treatment
>>>> programs (the grouping factor). The raw data are in
>>> miles driven from
>>>> the client's residence to the treatment program.
>>> A natural logarithm
>>>> transformation of miles driven works well to reduce
>>> the positive skew in
>>>> miles driven. I fit a model with lme() that looks
>>> like this:
>>>>
>>>> Linear mixed-effects model fit by REML
>>>> Data: dist.df
>>>> AIC BIC logLik
>>>> 38145.37 38319.54 -19049.68
>>>>
>>>> Random effects:
>>>> Formula: ~1 | PROGRAM
>>>> (Intercept) Residual
>>>> StdDev: 0.4268969 0.8988483
>>>>
>>>> Fixed effects: log(DIST5DZ + 1) ~ QUAD + BEAL_TRI +
>>> log(RZIPAREA + 1) +
>>>> log(PZIPAREA + 1) + AGE.TRI + P3GEND + P5RACEX + EMPLD
>>> + P13REASN +
>>>> METHFST + URGE.DI + WDRAW.DI + RX_30 + P7HR30
>>>>
>>>> Value Std.Error DF
>>> t-value p-value
>>>> (Intercept) 1.2235603 0.13153231 14288
>>> 9.30236 0.0000
>>>> QUADSouthEast 0.2100666 0.13200891 76
>>> 1.59131 0.1157
>>>> QUADMidWest 0.2760390 0.15709516 76
>>> 1.75715 0.0829
>>>> QUADWest -0.1655914 0.15536003 76
>>> -1.06586 0.2899
>>>> BEAL_TRI250K-1M -0.0264939 0.11724713 76
>>> -0.22597 0.8218
>>>> BEAL_TRI<250K -0.0965256 0.16399464
>>> 76 -0.58859 0.5579
>>>> log(RZIPAREA + 1) 0.2965304 0.00757138 14288
>>> 39.16463 0.0000
>>>> log(PZIPAREA + 1) -0.0042061 0.04413826 76
>>> -0.09529 0.9243
>>>> AGE.TRI30-43 -0.0309444 0.01789442 14288
>>> -1.72927 0.0838
>>>> AGE.TRI43-83 -0.1281177 0.02168648 14288
>>> -5.90772 0.0000
>>>> P3GENDFemale -0.0195289 0.01632703 14288
>>> -1.19611 0.2317
>>>> P5RACEXLatino -0.3527584 0.02904416 14288
>>> -12.14559 0.0000
>>>> P5RACEXBlack -0.5485861 0.03306146 14288
>>> -16.59292 0.0000
>>>> P5RACEXOther -0.1580669 0.04811350 14288
>>> -3.28529 0.0010
>>>> EMPLDYes 0.0098856 0.01650635 14288
>>> 0.59890 0.5493
>>>> P13REASNYes -0.0095057 0.01650128 14288
>>> -0.57606 0.5646
>>>> METHFSTYes 0.0073478 0.01672948 14288
>>> 0.43921 0.6605
>>>> URGE.DI Strong-VeryStrong -0.0272769 0.02400068 14288
>>> -1.13651 0.2558
>>>> WDRAW.DISevere-VerySevere 0.0012054 0.01810067 14288
>>> 0.06659 0.9469
>>>> RX_30Yes 0.0934411 0.02165389 14288
>>> 4.31521 0.0000
>>>> P7HR30Yes -0.0408189 0.02256842 14288
>>> -1.80867 0.0705
>>>>
>>>> Standardized Within-Group Residuals:
>>>> Min Q1 Med Q3
>>> Max
>>>> -4.17921640 -0.49875991 0.08672984 0.59020542
>>> 4.54644432
>>>>
>>>> Number of Observations: 14385
>>>> Number of Groups: 83
>>>>
>>>> I would like to summarize the fixed effects in terms
>>> of miles rather
>>>> than log(miles + 1). How can that be done? Are there
>>> common
>>>> generalized linear mixed models for miles driven that
>>> would avoid the
>>>> transformation and allow effects to be presented in
>>> miles?
>>>>
>>>> thanks,
>>>>
>>>> Chuck
>>>>
>>>
>>> --
>>> Dimitris Rizopoulos
>>> Assistant Professor
>>> Department of Biostatistics
>>> Erasmus University Medical Center
>>>
>>> Address: PO Box 2040, 3000 CA Rotterdam, the Netherlands
>>> Tel: +31/(0)10/7043478
>>> Fax: +31/(0)10/7043014
>>>
>>>
>>>
>>> ------------------------------
>>>
>>> _______________________________________________
>>> R-sig-mixed-models mailing list
>>> R-sig-mixed-models at r-project.org
>>> https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models
>>>
>>>
>>> End of R-sig-mixed-models Digest, Vol 27, Issue 36
>>> **************************************************
>>
>>
>> Yahoo! Cocina
>> Recetas prácticas y comida saludable
>> http://ar.mujer.yahoo.com/cocina/
>>
>>
>
>
> --
> The University of Edinburgh is a charitable body, registered in
> Scotland, with registration number SC005336.
>
>
>
>
> Yahoo! Cocina
> Recetas prácticas y comida saludable
> http://ar.mujer.yahoo.com/cocina/
>
>
--
The University of Edinburgh is a charitable body, registered in
Scotland, with registration number SC005336.
More information about the R-sig-mixed-models
mailing list