Hello, ? I am trying to apply linear models to some observational data.? The data are not balanced.? Basically, I am wish to describe sexual and individual signatures in calls of a bird species. I have many call recordings and measured several acoustic variables from each recording. Here's an analysis for one variable: ?
lmer(MH11~SEX+(1|BIRD)+(1|NIGHT), data=calls)
Linear mixed model fit by REML? Formula: MH11 ~ SEX + (1 | BIRD) + (1 | NIGHT)? ?? Data: calls? ? AIC? BIC logLik deviance REMLdev ?3662 3681? -1826???? 3663??? 3652 Random effects: ?Groups?? Name??????? Variance Std.Dev. ?NIGHT??? (Intercept)? 652.77? 25.549?? ?BIRD???? (Intercept) 1083.67? 32.919?? ?Residual????????????? 966.50? 31.089?? Number of obs: 356, groups: NIGHT, 138; BIRD, 57 ? Fixed effects: ??????????? Estimate Std. Error t value (Intercept)? 445.059????? 8.212?? 54.19 SEXM????????? 94.779???? 10.588??? 8.95 ? Correlation of Fixed Effects: ???? (Intr) SEXM -0.776 ? SEX is a fixed effect (Female/Male). BIRD (individuals) and NIGHT are random effects.? The nesting is SEX/BIRD/NIGHT/CALL. There is a clear difference between-SEXes (fixed effect), i.e. SEXF and SEX M have different voices. Secondly, there is some interesting variation between-BIRDs. Individual BIRDs have different voices. What if the BIRD random effect was not the same in both SEXF and SEXM groups? That would confound the model. Since there is a fixed effect (sexual signature), perhaps the random effect (individual signature) is different between groups SEXF and SEXM as well (unexpected, but should be tested). I'm not sure how to test for differences between groups in the random effects. One possibility is to split SEXF and SEXM and run separate lmer models. How would I the compare the results? Another is to compare the distributions of random effects between SEXF and SEXM (a two-sample test?). ? I would appreciate some advice on how to proceed? Stephen from Australia.