add nbinom2

R/compute_variances.R

@@ -675,7 +675,8 @@
       beta = mu,
       ordbeta = stats::plogis(mu),
       poisson = ,
-      nbinom1 = sqrt(exp(mu + 0.5 * as.vector(revar_null))),
+      nbinom1 = ,
+      nbinom2 = exp(mu + 0.5 * as.vector(revar_null)),
       exp(mu)
     )
   }
@@ -701,11 +702,11 @@
 
         # (zero-inflated) negative binomial ----
         # --------------------------------------
-        nbinom1 = sig,
+        nbinom1 = ,
+        nbinom2 = sig,
         `zero-inflated negative binomial` = ,
         `negative binomial` = ,
-        genpois = ,
-        nbinom2 = .variance_family_nbinom(x, mu, sig, faminfo),
+        genpois = .variance_family_nbinom(x, mu, sig, faminfo),
         truncated_nbinom2 = stats::family(x)$variance(mu, sig),
 
         # other distributions ----
@@ -726,7 +727,14 @@
           "Model's distribution-specific variance is negative. Results are not reliable."
         )
       }
-      vv / mu^2
+
+      # now compute cvsquared
+      switch(faminfo$family,
+        nbinom2 = (1 / mu) + (1 / sig),
+        poisson = ,
+        nbinom1 = vv / mu,
+        vv / mu^2
+      )
     },
     error = function(x) {
       if (verbose) {

tests/testthat/test-r2_nakagawa_MuMIn.R

@@ -385,7 +385,7 @@ test_that("lme4, Poisson", {
 
 
 # ==============================================================================
-# Poisson mixed models, glmmTMB
+# neg-binomial1 mixed models, glmmTMB
 # ==============================================================================
 
 test_that("glmmTMB, Nbinom1", {
@@ -428,3 +428,37 @@ test_that("glmmTMB, Nbinom1", {
   expect_equal(out1[1, "R2m"], out2$R2_marginal, ignore_attr = TRUE, tolerance = 1e-4)
   expect_equal(out1[1, "R2c"], out2$R2_conditional, ignore_attr = TRUE, tolerance = 1e-4)
 })
+
+
+# ==============================================================================
+# neg-binomial2 mixed models, glmmTMB
+# ==============================================================================
+
+test_that("glmmTMB, Nbinom2", {
+  # dataset ---------------------------------
+  data(Salamanders, package = "glmmTMB")
+
+  # no random slopes
+  m <- glmmTMB::glmmTMB(
+    count ~ mined + spp + (1 | site),
+    family = glmmTMB::nbinom2(),
+    data = Salamanders
+  )
+  out1 <- suppressWarnings(MuMIn::r.squaredGLMM(m))
+  out2 <- performance::r2_nakagawa(m)
+  # matches theoretical values
+  expect_equal(out1[2, "R2m"], out2$R2_marginal, ignore_attr = TRUE, tolerance = 1e-4)
+  expect_equal(out1[2, "R2c"], out2$R2_conditional, ignore_attr = TRUE, tolerance = 1e-4)
+
+  # with random slopes
+  m <- suppressWarnings(glmmTMB::glmmTMB(
+    count ~ mined + spp + cover + (1 + cover | site),
+    family = glmmTMB::nbinom2(),
+    data = Salamanders
+  ))
+  out1 <- suppressWarnings(MuMIn::r.squaredGLMM(m))
+  out2 <- performance::r2_nakagawa(m, tolerance = 1e-8)
+  # matches theoretical values
+  expect_equal(out1[2, "R2m"], out2$R2_marginal, ignore_attr = TRUE, tolerance = 1e-1)
+  expect_equal(out1[2, "R2c"], out2$R2_conditional, ignore_attr = TRUE, tolerance = 1e-1)
+})