attempt to find better starting values for GLMM when defaults aren't positive semidefinite (#796)

* re-initialize to small values

* simplify example

* BlueStyle

* use StableRNG

* test

* NEWS + version bump

* add zenodo JSON

Author: palday

.zenodo.json

@@ -0,0 +1,19 @@
+{
+    "access_right": "open",
+    "license": "MIT",
+    "title": "MixedModels.jl",
+    "description": "A Julia package for fitting (statistical) mixed-effects models",
+    "upload_type": "software",
+    "creators": [
+        {
+            "affiliation": "Beacon Biosignals",
+            "name": "Alday, Phillip M.",
+            "orcid": "0000-0002-9984-5745"
+        },
+        {
+            "affiliation": "University of Wisconsin - Madison",
+            "name": "Bates, Douglas",
+            "orcid": "0000-0001-8316-9503"
+        }
+    ]
+}

NEWS.md

@@ -1,3 +1,7 @@
+MixedModels v4.28.0 Release Notes
+==============================
+- `GeneralizedLinearMixedModel` now attempts to fall back to very constrained variance values when the default initial values result in a non positive semidefinite covariance matrix. [#792]
+
 MixedModels v4.27.1 Release Notes
 ==============================
 - `profile` now includes a `finally` block to restore the original model even if an error occurs before profiling is complete [#795]
@@ -579,4 +583,5 @@ Package dependencies
 [#783]: https://github.com/JuliaStats/MixedModels.jl/issues/783
 [#785]: https://github.com/JuliaStats/MixedModels.jl/issues/785
 [#791]: https://github.com/JuliaStats/MixedModels.jl/issues/791
+[#792]: https://github.com/JuliaStats/MixedModels.jl/issues/792
 [#795]: https://github.com/JuliaStats/MixedModels.jl/issues/795

Project.toml

@@ -1,7 +1,7 @@
 name = "MixedModels"
 uuid = "ff71e718-51f3-5ec2-a782-8ffcbfa3c316"
-author = ["Phillip Alday <[email protected]>", "Douglas Bates <[email protected]>", "Jose Bayoan Santiago Calderon <[email protected]>"]
-version = "4.27.1"
+author = ["Phillip Alday <[email protected]>", "Douglas Bates <[email protected]>"]
+version = "4.28.0"
 
 [deps]
 Arrow = "69666777-d1a9-59fb-9406-91d4454c9d45"