Skip to content

What p value should I report here?

2 messages · DESPINA MICHAILIDOU, Ben Bolker

Original
DESPINA MICHAILIDOU
# 
Hi everyone,


I am running this regression analysis model and I get the following output.
What P value should I report for my variable P-Dizz today?What does it mean
that fixed-effect model matrix is rank deficient so dropping 1 column /
coefficient? Can anyone help me with the interpretation of those data?


Thank you in advance.


Despina


Generalized linear mixed model fit by maximum likelihood (Laplace
Approximation) ['glmerMod']

 Family: binomial  ( logit )

Formula: Vert_effect ~ P_Diz_today + (1 | ID/SCAN_DATE/Side)

   Data: GCA_data



     AIC      BIC   logLik deviance df.resid

    80.3     94.5    -36.1     72.3      254



Scaled residuals:

      Min        1Q    Median        3Q       Max

-0.012501 -0.000639 -0.000639 -0.000639  0.105723



Random effects:

 Groups              Name        Variance Std.Dev.

 Side:(SCAN_DATE:ID) (Intercept) 1502.7   38.76

 SCAN_DATE:ID        (Intercept)    0.0    0.00

 ID                  (Intercept)  235.1   15.33

Number of obs: 258, groups:  Side:(SCAN_DATE:ID), 258; SCAN_DATE:ID, 130;
ID, 52



Fixed effects:

            Estimate Std. Error z value Pr(>|z|)

(Intercept)  -14.711      3.646  -4.035 5.47e-05 ***

---

Signif. codes:  0 ?***? 0.001 ?**? 0.01 ?*? 0.05 ?.? 0.1 ? ? 1

fit warnings:

fixed-effect model matrix is rank deficient so dropping 1 column /
coefficient

convergence code: 0

boundary (singular) fit: see ?isSingular
Ben Bolker
# 
Can you show us summary(GCA_data) and
summary(model.frame(fitted_model)) please? It looks like for some reason
(maybe because of observations dropped due to NA values?) you have no
variation in your predictor variable (P_Diz_today).

  It's also potentially problematic that you have an observation-level
random effect for a Bernoulli outcome (i.e., you're fitting a binomial
model with a single-column value as the response and no weights=
argument, which implies you have a 0/1 outcome; you have the same number
of groups in your fully nested [ID:Scan:Side] random effect as
observations), but I don't think this would lead to the dropping of the
P_Diz_today predictor ...

  cheers
    Ben Bolker
On 2019-05-02 3:30 p.m., DESPINA MICHAILIDOU wrote: