At my wits end running models and would greatly appreciate help. I am currently running GLMM's to understand the relationships between a set of predictor variables (continuous data) and my binomial outcome variable (0,1 - 'self_reported_conflict). Ultimately, my goal is to determine what variables (or set therof) predict conflict outcomes between people and bears.
I have data about predictor/outcome variables for 201 households in my study area. I'd like to run log link function GLMMs with a random effect for 'household'. I have 194 households and 201 observations (most households have 1 observation, several have 2). However, whenever I include my random effect (household) only in my model, my fit is singular (boundary (singular) fit: see help('isSingular').
I can't for the life of me understand why this is occuring in my most basic model (random effect only). I've read other posts with similar questions but can't seem to extrapolate the reasons to my data.
Any help immensely appreciate.
I try the following code for my null model (with random effect):
>m.e1 <- glmer(self_reported_conflict ~ 1 + (1|household),
data = df,
family = binomial(link = "logit"))
>summary(m.e1)
Generalized linear mixed model fit by maximum likelihood (Laplace
Approximation) [glmerMod]
Family: binomial ( logit )
Formula: self_reported_conflict ~ 1 + (1 | household)
Data: df
AIC BIC logLik deviance df.resid
213.6 220.3 -104.8 209.6 201
Scaled residuals:
Min 1Q Median 3Q Max
-0.5184 -0.5184 -0.5184 -0.5184 1.9290
Random effects:
Groups Name Variance Std.Dev.
household (Intercept) 0 0
Number of obs: 203, groups: household, 194
Fixed effects:
Estimate Std. Error z value Pr(>|z|)
(Intercept) -1.3140 0.1718 -7.649 2.02e-14 ***
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
optimizer (Nelder_Mead) convergence code: 0 (OK)
boundary (singular) fit: see help('isSingular')'
