[R-sig-ME] lmer or glmer?

[ONKELINX, Thierry]

Dear Michael,

Neither of the qq plots look terribly problematic. But don't just look at qq plots! Plot the residuals against the available covariates and see if there is any pattern.

Choose a distribution family based on the properties of the response. Poisson assumes non-negative integer values. So don't use Poisson if the response is continuous. You will need to tell us more about the response if you need help on that.

Best regards,

ir. Thierry Onkelinx
Instituut voor natuur- en bosonderzoek / Research Institute for Nature and Forest
team Biometrie & Kwaliteitszorg / team Biometrics & Quality Assurance
Kliniekstraat 25
1070 Anderlecht
Belgium
+ 32 2 525 02 51
+ 32 54 43 61 85
Thierry.Onkelinx op inbo.be
www.inbo.be
To call in the statistician after the experiment is done may be no more than asking him to perform a post-mortem examination: he may be able to say what the experiment died of. ~ Sir Ronald Aylmer Fisher
The plural of anecdote is not data. ~ Roger Brinner
The combination of some data and an aching desire for an answer does not ensure that a reasonable answer can be extracted from a given body of data. ~ John Tukey

________________________________________
Van: R-sig-mixed-models [r-sig-mixed-models-bounces op r-project.org] namens Michael Jackson [Michael.Jackson op vuw.ac.nz]
Verzonden: maandag 12 januari 2015 20:41
Aan: r-sig-mixed-models op r-project.org
Onderwerp: Re: [R-sig-ME] lmer or glmer?

Hi Thierry,

Thanks for the input. Ive attached some links to qqnorm(resid(x)) plots run for varying data.

Plot 1 (link below) (based on lmer code) is characteristic of all my plots when using my raw data, i.e., showing positive skew to a greater or lesser extent than this example. I identified the response variable as key to driving this and therefore tried log transforming my response variable.

Plot 2 is the same raw data and lmer coding, but with a log transformed response variable. It now shows slight negative skew and if anything is worse.

Plot 3 is from my glmer coding I proposed in my first message (the two former used the lmer coding) and also my raw untransformed response or predictors. This looks lots better. A SW test is also not significant (p=0.67). This is actually based on using "family=poisson(link=log)" as ive now read that the "family=" aspect only relates to the response, not predictors ...Id be interested in your thoughts.


Plot 1 http://s166.photobucket.com/user/michaeljackson1972/media/Plot1_zps1e8eb440.png.html

Plot 2 http://s166.photobucket.com/user/michaeljackson1972/media/Plot2_zpsa5e555b8.png.html

Plot 3 http://s166.photobucket.com/user/michaeljackson1972/media/Plot3_zpsf4a5b638.png.html

Thanks again,

Mike


..............
PhD Candidate
Centre for Biodiversity and Restoration Ecology
Room KK 411
Kirk Building
Kelburn Parade
Wellington 6012

        [[alternative HTML version deleted]]

_______________________________________________
R-sig-mixed-models op r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models
Disclaimer<https://www.inbo.be/nl/disclaimer-mailberichten-van-het-inbo>