edits to vignette; re-build pkgdown

friendly · Sep 20, 2022 · de77bff · de77bff
1 parent ecafa87
commit de77bff
Show file tree

Hide file tree

Showing 9 changed files with 83 additions and 8 deletions.
diff --git a/docs/articles/uni-vs-multi.html b/docs/articles/uni-vs-multi.html
diff --git a/docs/articles/uni-vs-multi_files/figure-html/dfbetas-plot-1.png b/docs/articles/uni-vs-multi_files/figure-html/dfbetas-plot-1.png
diff --git a/docs/articles/uni-vs-multi_files/figure-html/inflplot-mlm-1.png b/docs/articles/uni-vs-multi_files/figure-html/inflplot-mlm-1.png
diff --git a/docs/articles/uni-vs-multi_files/figure-html/inflplot-mlm-lr-1.png b/docs/articles/uni-vs-multi_files/figure-html/inflplot-mlm-lr-1.png
diff --git a/docs/articles/uni-vs-multi_files/figure-html/inflplots-1.png b/docs/articles/uni-vs-multi_files/figure-html/inflplots-1.png
diff --git a/docs/articles/uni-vs-multi_files/figure-html/inflplots-2.png b/docs/articles/uni-vs-multi_files/figure-html/inflplots-2.png
diff --git a/docs/articles/uni-vs-multi_files/figure-html/scatmat-1.png b/docs/articles/uni-vs-multi_files/figure-html/scatmat-1.png
diff --git a/extra/Toy.R b/extra/Toy.R
@@ -38,7 +38,19 @@ coef(Toy.lm1)
 coef(Toy.lm2)
 coef(Toy.mlm)
 
+# or, using broom::tidy
+broom::tidy(Toy.lm1)
+broom::tidy(Toy.lm2)
+broom::tidy(Toy.mlm)
 
+broom::glance(Toy.lm1)
+broom::glance(Toy.lm2)
+# But there is no glance method for an mlm
+# broom::glance(Toy.mlm)
+
+
+
+# Cook's distance
 df <- Toy
 df$D1  <- cooks.distance(Toy.lm1)
 df$D2  <- cooks.distance(Toy.lm2)

diff --git a/vignettes/uni-vs-multi.Rmd b/vignettes/uni-vs-multi.Rmd
@@ -98,6 +98,25 @@ Toy.lm2 <- lm(y2 ~ x, data=Toy)
 Toy.mlm <- lm(cbind(y1, y2) ~ x, data=Toy)
 ```
 
+Note that the coefficients in the multivariate model `Toy.mlm` are identical to those in the separate
+univariate models for `y1` and `y2`. That is, 
+$\mathbf{B} = [ \mathbf{b_{y1}} , \mathbf{b_{y2}} ]$, as if the univariate models were fit individually.
+
+```{r coefs}
+coef(Toy.lm1)
+coef(Toy.lm2)
+coef(Toy.mlm)
+```
+
+However, the test for predictors differ, because the multivariate tests take the correlation between `y1` and `y2` into account.
+
+```{r anovas}
+car::Anova(Toy.lm1)
+car::Anova(Toy.lm2)
+car::Anova(Toy.mlm)
+```
+
+
 ### Cook's D
 
 First, let's examine the Cook's D statistics for the models. Note that the function `cooks.distance()`