[R-sig-ME] Questions regarding bs() interaction in lmer()

[Ben Bolker]

On 2022-05-27 2:17 p.m., Hedyeh Ahmadi wrote:
> Hello All,
> I am running a lmer() model as follows. The following model gives all the interactions for before and after knots shown below. I have 3 questions:
> 
>    1.  I would like to suppress the highlighted interactions that is before and after the knot. I have tried the update() command with -bs(reshist, knots = 7, degree = 1)2:bs(age, knots = 126, degree = 1)1, which did not work.

   It's not clear to me that there's an easy way to suppress particular 
components of such an interaction.

>    2.  How would I define the knots for the interaction term and is what I have a correct specification?

   The knots would be constructed from the interaction.

>    3.  Is there a package that would take this model and output interpretable parameter estimates for splines?

   I don't really know of *any* spline framework that gives readily 
interpretable parameter estimates.  In my experience people usually look 
at a variety of different kinds of prediction plots.

    If you're going to do a lot of stuff with splines I would recommend 
the mgcv package, which allows a broad range of flexible spline 
parameterizations, along with the 'gratia' package for post-processing 
and the paper by Pedersen et al on hierarchical GAMs:

Pedersen, Eric J., David L. Miller, Gavin L. Simpson, and Noam Ross. 
“Hierarchical Generalized Additive Models in Ecology: An Introduction 
with Mgcv.” PeerJ 7 (May 27, 2019): e6876. 
https://doi.org/10.7717/peerj.6876.

   You might get more useful advice from the list if you give a little 
bit more context / reasons *why* you want to do these particular things ...

   cheers
    Ben Bolker
> 
> Thank you in advance for your time
> 
> smri_left <-
>    lmer(smri ~ 1 +
>           bs(reshist, knots=7, degree = 1)+
>           bs(age, knots=126, degree = 1)+
>           bs(reshist, knots=7.787, degree = 1)*bs(age, knots=126, degree = 1)+
>           ethnicity.1_bl + high.bl + neighb_avg_p_bl
>           (1|site/subject)  ,
>           REML = FALSE,
>           control = lmerControl(optimizer = "Nelder_Mead"),
>           data=WG)
> 
> Fixed effects:
>                                                                                                                              Estimate         Std. Error     df                      t value             Pr(>|t|)
> (Intercept)                                                                                                         2.6679353    0.0114634  652.5132635    232.736        < 0.0000000000000002 ***
> bs(reshist, knots = 7, degree = 1)1                                                                0.0090600    0.0100475 1059.5955954   0.902              0.36741
> bs(reshist, knots = 7, degree = 1)2                                                                0.0082266    0.0132667  559.1847547   0.620              0.53545
> bs(age, knots = 126, degree = 1)1                                                                -0.0137868    0.0081986 6143.2470351  -1.682              0.09270 .
> bs(age, knots = 126, degree = 1)2                                                                -0.0834209    0.0094686 5534.8186306  -8.810          < 0.0000000000000002 ***
> ethnicity.1_blHispanic                                                                                     0.0154627    0.0030354 3492.3232613   5.094           0.00000036893 ***
> ethnicity.1_blOther                                                                                          0.0158286    0.0033031 8304.5129245   4.792             0.00000167908 ***
> high.blBachelor                                                                                                 0.0022701    0.0044227 8691.0032954   0.513              0.60777
> high.blHS Diploma/GED                                                                                  -0.0009402    0.0045056 8757.0427133  -0.209              0.83472
> neighb_avg_p_bl                                                                                              0.0001371    0.0009128 8628.5294654   0.150              0.88060
> bs(reshist, knots = 7, degree = 1)1:bs(age, knots = 126, degree = 1)1   -0.0091887    0.0095058 6161.8339520  -0.967              0.33376
> bs(reshist, knots = 7, degree = 1)2:bs(age, knots = 126, degree = 1)1   -0.0148081    0.0128074 6412.6567741  -1.156              0.24764
> bs(reshist, knots = 7, degree = 1)1:bs(age, knots = 126, degree = 1)2    0.0080632    0.0109929 5538.9731189   0.733              0.46329
> bs(reshist, knots = 7, degree = 1)2:bs(age, knots = 126, degree = 1)2   -0.0592144    0.0150237 5595.8147976  -3.941        0.00008199861 ***
> ---
> Signif. codes:  0 �***� 0.001 �**� 0.01 �*� 0.05 �.� 0.1 � � 1
> 
> Best,
> 
> Hedyeh Ahmadi, Ph.D.
> Statistician
> Keck School of Medicine
> Department of Preventive Medicine
> University of Southern California
> 
> LinkedIn
> www.linkedin.com/in/hedyeh-ahmadi<http://www.linkedin.com/in/hedyeh-ahmadi>
> <http://www.linkedin.com/in/hedyeh-ahmadi><http://www.linkedin.com/in/hedyeh-ahmadi>
> 
> 
> 
> 
> 
> 	[[alternative HTML version deleted]]
> 
> 
> _______________________________________________
> R-sig-mixed-models using r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models

-- 
Dr. Benjamin Bolker
Professor, Mathematics & Statistics and Biology, McMaster University
Director, School of Computational Science and Engineering
(Acting) Graduate chair, Mathematics & Statistics