[R-sig-ME] LME4: output interpretation of tricky model

ONKELINX, Thierry Thierry.ONKELINX at inbo.be
Tue May 22 12:48:26 CEST 2012


Dear Eiko,

Try to write out the model by hand. E.g. if you want a prediction for Y_index = 1, time = 1, x1 = 2 and x2 = 4, which coefficient would you use? And what is the values for Y_index, time, x1 and x2 change (look especially at time = 0, x1 =  and x2 = 0)?

If the interpretation of the coefficients is important, then I would model it as

Y ~ -1 + time + Y_index:x1 + Y_index:x2 + ( - 1 + Y_index | subject)

Which gives the same model fit but with a different parametrisation.

Best regards,

Thierry

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


-----Oorspronkelijk bericht-----
Van: r-sig-mixed-models-bounces op r-project.org [mailto:r-sig-mixed-models-bounces op r-project.org] Namens Eiko Fried
Verzonden: zondag 20 mei 2012 13:53
Aan: r-sig-mixed-models op r-project.org
Onderwerp: [R-sig-ME] LME4: output interpretation of tricky model

Dear Mailinglist,

I would be very glad to get some assistance with interpreting a tricky model output in LME4.

The model is this:

Y ~ -1 + Y_index + time + Y_index*x1 + Y_index*x2 + ( - 1 + Y_index |
subject)

What I am doing with this is modeling 9 items of a questionnaire as a multivariate response variable Y.
Y_index is a categorical variable defining the number of item of the questionnaire (1 through 9), and I am checking with interaction effects on
x1 and x2 whether these covariates have differential effects on my multivariate response. It is a longitudinal design with 5 measurement points, and I expect that x1 only affects some of my 9 items (the same for x2).

The dataset is in long-long format (Y_index * time), so each subject has
9*5 lines.
(I found the suggestion for that kind of analysis in Hox, 2010).

The for me relevant part of the output looks like this:

Fixed effects:
                              Estimate    Std. Error    t value
Y_index1                 0.3161592  0.0780922   4.049
Y_index2                 0.4685218  0.0775360   6.043
Y_index3                 0.9531528  0.0969119   9.835
Y_index4                 0.2366093  0.0898923   2.632
Y_index5                 0.3055025  0.0955639   3.197
Y_index6                 0.2581729  0.0819606   3.150
Y_index7                 0.4556287  0.0817002   5.577
Y_index8                 0.6027990  0.0691566   8.716
Y_index9                 0.8697155  0.0620898  14.007
time                        0.5726978  0.0374384  15.297
x1                           0.0196260  0.0020225   9.704
x2                          -0.0415874  0.0350631  -1.186
Y_index2:x1            0.0023080  0.0018770   1.230
Y_index3:x1           -0.0019870  0.0027166  -0.731
Y_index4:x1           -0.0006285  0.0023784  -0.264
Y_index5:x1            0.0033737  0.0026178   1.289
Y_index6:x1            0.0067164  0.0020428   3.288
Y_index7:x1           -0.0016510  0.0021435  -0.770
Y_index8:x1           -0.0080817  0.0020414  -3.959
Y_index9:x1           -0.0140743  0.0021874  -6.434
Y_index2:x2            0.0358944  0.0325015   1.104
Y_index3:x2           -0.0675878  0.0470604  -1.436
Y_index4:x2            0.0037518  0.0411980   0.091
Y_index5:x2           -0.0456199  0.0453805  -1.005
Y_index6:x2            0.0333067  0.0353716   0.942
Y_index7:x2            0.0443440  0.0371040   1.195
Y_index8:x2            0.0595453  0.0353532   1.684
Y_index9:x2            0.0223958  0.0379038   0.591

What I don't understand is
(1) why Y_index1 is missing in my interaction output with x1 and x2 (the lists start with Y_index2), and
(2) what the interaction lines exactly mean. Is it a comparison to a baseline? Is it difference from the main effect?

I do apologize in case this is a very simple question, but I cannot get my head around it.

Thank you
--E

        [[alternative HTML version deleted]]

_______________________________________________
R-sig-mixed-models op r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models
* * * * * * * * * * * * * D I S C L A I M E R * * * * * * * * * * * * *
Dit bericht en eventuele bijlagen geven enkel de visie van de schrijver weer en binden het INBO onder geen enkel beding, zolang dit bericht niet bevestigd is door een geldig ondertekend document.
The views expressed in this message and any annex are purely those of the writer and may not be regarded as stating an official position of INBO, as long as the message is not confirmed by a duly signed document.



More information about the R-sig-mixed-models mailing list