recommended exactLRT for boundary space sigma

Author: mahort

Lecture-6-BroadSenseHeritability.ipynb

@@ -598,8 +598,11 @@
     "\n",
     "        ourEstimate <- r.squared.merMod(lmer.alt);\n",
     "        results[nrow(results), \"H2\"] <- ourEstimate['Conditional'] - ourEstimate['Marginal'];\n",
-    "        results[nrow(results), \"pvalue\"] <- anova(lmer.alt, lm.null)[2, \"Pr(>Chisq)\"];\n",
     "\n",
+    "        ## note that the mixed model has to be specified first\n",
+    "        ## if sigma is on the boundary, this approach is conservative; consider exactLRT which estimates the p-value using simulations.\n",
+    "        ## source: https://bbolker.github.io/mixedmodels-misc/glmmFAQ.html\n",
+    "        results[nrow(results), \"pvalue\"] <- anova(lmer.alt, lm.null)[2, \"Pr(>Chisq)\"];\n",
     "}"
    ]
   },
@@ -727,7 +730,7 @@
     }
    ],
    "source": [
-    "results"
+    "results;"
    ]
   },
   {