caption-colors

06-linear_models-plots.qmd

@@ -863,13 +863,16 @@ in the fitted values for the term being graphed.
 In practical terms, a scoring matrix $\mathbf{X}^\bullet$ is defined by
 the focal variables varied over their ranges and the other variables held
 fixed.  The fitted values for a model term are then calculated as 
-$\widehat{\mathbf{y}}^\bullet = \mathbf{X}^\bullet \hat{\mathbf{b}}$
+$\widehat{\mathbf{y}}^\bullet = \mathbf{X}^\bullet \; \widehat{\mathbf{b}}$
 using the equivalent of:
 ```{r}
 #| eval: false
 predict(model, newdata = X, se.fit = TRUE) 
 ```
-which also calculates the standard errors. 
+which also calculates the standard errors
+as the square roots of
+$\diag (\mathbf{X}^\bullet \, \widehat{\mathbf{\Var}} (\mathbf{b}) \, \mathbf{X}^{\bullet\mathsf{T}} )$
+where $\widehat{\mathbf{\Var}} (\mathbf{b})$ is the estimated covariance matrix of the coefficients.
 Consequently, predictor effect values can be obtained for _any_ modelling function that has
 `predict()` and `vcov()` methods. To date, effect displays are available for
 over 100 different model types in various packages.
@@ -1058,7 +1061,7 @@ predictorEffects(prestige.mod3, ~ income,
 This provides a clear interpretation of the interaction, represented by the coefficients shown above for
 `log10(income):typewc` and `log10(income):typeprof` in the model. For all three occupation types,
 prestige increases linearly with log income, meaning that increasing income by 10% (say)
-gives and increase of $40.33 / 10 = 4.033$ in prestige. The slope for professional workers
+gives an increase of $40.33 / 10 = 4.033$ in prestige. The slope for professional workers
 is less steep, suggesting that for these workers, prestige increases less with multiples of income.
 
 

test/caption-colors.qmd

@@ -0,0 +1,128 @@
+---
+title: "How to use colored text in Quarto figure captions"
+format:
+  html:
+    toc: true
+  pdf:
+    toc: true
+---
+
+In figure captions, it is often useful to say refer to colored objects by printing
+the words in their colors.
+How can I do this with output in HTML or PDF?
+
+Define a function to colorize text, that can be used in HTML or PDF output
+
+```{r colorize}
+# colorize text: use inline as `r colorize(text, color)` to print `text` in a given `color`
+# can also be used to color a color name, as in r colorize("red")`
+colorize <- function(text, color) {
+  if (missing(color)) color <- text
+  if (knitr::is_latex_output()) {
+    sprintf("\\textcolor{%s}{%s}", color, text)
+  } else if (knitr::is_html_output()) {
+    sprintf("<span style='color: %s;'>%s</span>", color, text)
+  } else text
+}
+```
+
+In normal text, I can use `colorize()` to
+print text in `r colorize("red")`, `r colorize("blue")`, `r colorize("green")`, ... 
+
+But not in a figure caption.
+
+This works but the text is not colored in the caption of @fig-test
+
+````qmd
+#| label: fig-test
+#| out-width: "70%"
+#| fig-cap: Some points should be red, some blue, some green
+cols <- c("red", "blue", "green")
+plot(1:10, pch=16, cex = 3, col = cols)
+````
+
+```{r}
+#| label: fig-test
+#| out-width: "70%"
+#| fig-cap: Some points should be red, some blue, some green
+cols <- c("red", "blue", "green")
+plot(1:10, pch=16, cex = 3, col = cols)
+```
+
+## Inline expression in a caption
+`colorize()` doesn't work in a caption. The expression is not evaluated, and just prints in the caption
+as a literal, in code font.
+
+````qmd
+#| label: fig-test2
+#| out-width: "70%"
+#| fig-cap: Some points are `r colorize("red")`, some are `r colorize("blue")`, some are `r colorize("green")`
+cols <- c("red", "blue", "green")
+plot(1:10, pch=16, cex = 3, col = cols)
+````
+
+```{r}
+#| label: fig-test2
+#| out-width: "70%"
+#| fig-cap: Some points are `r colorize("red")`, some are `r colorize("blue")`, some are `r colorize("green")`
+cols <- c("red", "blue", "green")
+plot(1:10, pch=16, cex = 3, col = cols)
+```
+
+## Try !expr
+One solution: use the YAML `!expr` "tag" literal in `fig-cap`:
+
+````qmd
+#| label: fig-test3
+#| out-width: "70%"
+#| fig-cap: !expr paste("Some points are", colorize('red'), "some are", colorize('blue'), "some are", colorize("green"))
+cols <- c("red", "blue", "green")
+plot(1:10, pch=16, cex = 3, col = cols)
+````
+
+
+```{r}
+#| label: fig-test3
+#| out-width: "70%"
+#| fig-cap: !expr paste("Some points are", colorize('red'), "some are", colorize('blue'), "some are", colorize("green"))
+cols <- c("red", "blue", "green")
+plot(1:10, pch=16, cex = 3, col = cols)
+```
+
+## Can we simplify this?
+
+Define color names as global variables and use `glue::glue()`
+```{r}
+red <- colorize('red')
+blue <- colorize('blue')
+green <- colorize("green")
+library(glue)
+```
+
+````qmd
+#| label: fig-test4
+#| out-width: "70%"
+#| fig-cap: !expr glue("Some points are ", {red}, " some are ", {blue}, "some are ", {green})
+cols <- c("red", "blue", "green")
+plot(1:10, pch=16, cex = 3, col = cols)
+````
+Voila! This is much easier to type in captions, but need to add spaces
+
+```{r}
+#| label: fig-test4
+#| out-width: "70%"
+#| fig-cap: !expr glue("Some points are ", {red}, " some are ", {blue}, " some are ", {green})
+plot(1:10, pch=16, cex = 3, col = cols)
+```
+
+## Final solution
+
+Use the color name variables, but just use `paste()`
+
+````qmd
+#| label: fig-test4
+#| out-width: "70%"
+#| fig-cap: !expr paste("Some points are", red, "some are", blue, "some are", green)
+plot(1:10, pch=16, cex = 3, col = cols)
+````
+