Address two minor issues in #65

carpentries-incubator · Jun 6, 2022 · bfb3af3 · bfb3af3
1 parent 28d82fc
commit bfb3af3
Show file tree

Hide file tree

Showing 24 changed files with 96 additions and 22 deletions.
diff --git a/_episodes/01-introduction-to-high-dimensional-data.md b/_episodes/01-introduction-to-high-dimensional-data.md
@@ -445,6 +445,14 @@ library("minfi")
 {: .language-r}
 
 
+
+~~~
+Warning: replacing previous import 'utils::download.file' by
+'restfulr::download.file' when loading 'rtracklayer'
+~~~
+{: .warning}
+
+
 ~~~
 browseVignettes("minfi")
 ~~~

diff --git a/_episodes/02-high-dimensional-regression.md b/_episodes/02-high-dimensional-regression.md
@@ -48,6 +48,20 @@ Let's read in the data for this episode:
 ~~~
 library("here")
 library("minfi")
+~~~
+{: .language-r}
+
+
+
+~~~
+Warning: replacing previous import 'utils::download.file' by
+'restfulr::download.file' when loading 'rtracklayer'
+~~~
+{: .warning}
+
+
+
+~~~
 methylation <- readRDS(here("data/methylation.rds"))
 ~~~
 {: .language-r}

diff --git a/_episodes/03-regression-regularisation.md b/_episodes/03-regression-regularisation.md
@@ -56,6 +56,20 @@ First, let's read in the data from the last lesson.
 ~~~
 library("here")
 library("minfi")
+~~~
+{: .language-r}
+
+
+
+~~~
+Warning: replacing previous import 'utils::download.file' by
+'restfulr::download.file' when loading 'rtracklayer'
+~~~
+{: .warning}
+
+
+
+~~~
 methylation <- readRDS(here("data/methylation.rds"))
 
 ## here, we transpose the matrix to have features as rows and samples as columns

diff --git a/_episodes/04-regression-feature-selection.md b/_episodes/04-regression-feature-selection.md
@@ -60,6 +60,20 @@ possible combination of features to find the best one.
 ~~~
 library("here")
 library("minfi")
+~~~
+{: .language-r}
+
+
+
+~~~
+Warning: replacing previous import 'utils::download.file' by
+'restfulr::download.file' when loading 'rtracklayer'
+~~~
+{: .warning}
+
+
+
+~~~
 methylation <- readRDS(here("data/methylation.rds"))
 
 ## here, we transpose the matrix to have features as rows and samples as columns

diff --git a/_episodes/08-k-means.md b/_episodes/08-k-means.md
@@ -253,12 +253,13 @@ between points in the clusters.
 library("cluster")
 dist_mat <- dist(pcs)
 sil <- silhouette(cluster$cluster, dist = dist_mat)
-plot(sil)
+plot(sil, border = NA)
 ~~~
 {: .language-r}
 
 <img src="../fig/rmd-08-silhouette-1.png" title="plot of chunk silhouette" alt="plot of chunk silhouette" width="432" style="display: block; margin: auto;" />
 
+
 Let's plot the silhouette score on the original dimensions used to cluster
 the data. Here, we're mapping cluster membership to point shape, and silhouette
 width to colour.
@@ -337,7 +338,7 @@ belong to.
 > > 
 > > 
 > > ~~~
-> > plot(sil5)
+> > plot(sil5, border = NA)
 > > ~~~
 > > {: .language-r}
 > > 

diff --git a/_episodes/09-hierarchical.md b/_episodes/09-hierarchical.md
@@ -101,6 +101,20 @@ with in the regression lessons. Let's load the data and look at it.
 
 ~~~
 library("minfi")
+~~~
+{: .language-r}
+
+
+
+~~~
+Warning: replacing previous import 'utils::download.file' by
+'restfulr::download.file' when loading 'rtracklayer'
+~~~
+{: .warning}
+
+
+
+~~~
 library("here")
 library("ComplexHeatmap")
 

diff --git a/_episodes/10-mixture-models.md b/_episodes/10-mixture-models.md
@@ -95,7 +95,11 @@ deviation) that best fit the data.
 set.seed(66)
 univar <- rnorm(200)
 library("MASS")
-opt <- fitdistr(x = univar, densfun = dnorm, start = list(mean = 1, sd = 1))
+opt <- fitdistr(
+    x = univar,
+    densfun = dnorm,
+    start = list(mean = 1, sd = 1)
+)
 ~~~
 {: .language-r}
 
@@ -111,6 +115,9 @@ Warning in densfun(x, parm[1], parm[2], ...): NaNs produced
 
 
 ~~~
+## fitdistr here produces warnings because it does not constrain sd to be > 0.
+## we could stop this by supplying a lower bound in the `lower` argument
+## but it doesn't make any difference.
 fitted_mean <- opt$estimate[["mean"]]
 fitted_sd <- opt$estimate[["sd"]]
 hist(univar, freq = FALSE, breaks = "FD")
@@ -137,7 +144,8 @@ curve(
 >    *Hint: try `dlnorm` with parameter names `meanlog` and `sdlog`.*
 > 
 > > ## Solution
-> > 1. 
+> > 1. The estimates are very close to the empirical mean and SD estimates, as we
+> >    might expect.
 > >    
 > >    ~~~
 > >    opt
@@ -184,7 +192,8 @@ curve(
 > >    
 > >    ~~~
 > >    univar_exp <- exp(univar)
-> >    opt_exp <- fitdistr(x = univar_exp, densfun = dlnorm, start = list(meanlog = 1, sdlog = 1))
+> >    opt_exp <- fitdistr(x = univar_exp, densfun = dlnorm,
+> >        start = list(meanlog = 1, sdlog = 1))
 > >    ~~~
 > >    {: .language-r}
 > >    
@@ -200,6 +209,7 @@ curve(
 > >    
 > >    
 > >    ~~~
+> >    ## again, fitdistr produces warnings because it does not constrain sdlog to be > 0.
 > >    opt_exp
 > >    ~~~
 > >    {: .language-r}
@@ -273,12 +283,6 @@ It's important to note that we don't necessarily have to pick good starting
 values here, though it may help. You can see that below our initial starting
 "guess" is really bad in this case:
 
-
-~~~
-Warning: Removed 4 rows containing missing values (geom_bar).
-~~~
-{: .warning}
-
 <img src="../fig/rmd-10-mixture-animation-1.png" title="Alt" alt="Alt" width="432" style="display: block; margin: auto;" />
 
 We then assign each data point to the component that fits them better (this is
@@ -289,12 +293,6 @@ until the algorithm converges -- meaning that the components don't change from
 iteration to iteration. In this simple example, the algorithm converges after 
 one iteration, but this won't usually be the case!
 
-
-~~~
-Warning: Removed 4 rows containing missing values (geom_bar).
-~~~
-{: .warning}
-
 <img src="../fig/rmd-10-mix-converged-1.png" title="Alt" alt="Alt" width="432" style="display: block; margin: auto;" />
 
 The figures shown here were made manually, to be able to step through the 

diff --git a/_episodes_rmd/10-mixture-models.Rmd b/_episodes_rmd/10-mixture-models.Rmd
@@ -88,7 +88,14 @@ deviation) that best fit the data.
 set.seed(66)
 univar <- rnorm(200)
 library("MASS")
-opt <- fitdistr(x = univar, densfun = dnorm, start = list(mean = 1, sd = 1))
+opt <- fitdistr(
+    x = univar,
+    densfun = dnorm,
+    start = list(mean = 1, sd = 1)
+)
+## fitdistr here produces warnings because it does not constrain sd to be > 0.
+## we could stop this by supplying a lower bound in the `lower` argument
+## but it doesn't make any difference.
 fitted_mean <- opt$estimate[["mean"]]
 fitted_sd <- opt$estimate[["sd"]]
 hist(univar, freq = FALSE, breaks = "FD")
@@ -123,7 +130,8 @@ ggplot(data.frame(x = univar)) +
 >    *Hint: try `dlnorm` with parameter names `meanlog` and `sdlog`.*
 > 
 > > ## Solution
-> > 1. 
+> > 1. The estimates are very close to the empirical mean and SD estimates, as we
+> >    might expect.
 > >    ```{r stats-univar}
 > >    opt
 > >    mean(univar)
@@ -132,7 +140,9 @@ ggplot(data.frame(x = univar)) +
 > > 2. 
 > >    ```{r fit-dlnorm}
 > >    univar_exp <- exp(univar)
-> >    opt_exp <- fitdistr(x = univar_exp, densfun = dlnorm, start = list(meanlog = 1, sdlog = 1))
+> >    opt_exp <- fitdistr(x = univar_exp, densfun = dlnorm,
+> >        start = list(meanlog = 1, sdlog = 1))
+> >    ## again, fitdistr produces warnings because it does not constrain sdlog to be > 0.
 > >    opt_exp
 > >    mean(univar_exp)
 > >    sd(univar_exp)
@@ -194,6 +204,7 @@ plot_mixture <- function(x, means, sds, components) {
         aes(x) +
         geom_histogram(
             aes(y = ..density.., fill = factor(paste("component", components))),
+            na.rm = TRUE,
             alpha = 0.4
         ) +
         lims(x = range(x) * 1.5) +
@@ -370,14 +381,14 @@ both variables at once. In fact, in some cases it can basically be identical:
 
 ```{r mvnorm, echo=FALSE}
 xy <- mvrnorm(1000, mu  = c(0, 0), Sigma = diag(2))
-plot(xy)
+plot(xy, xlab="x", ylab="y")
 ```
 
 However, it also allows us to model sets of variables that aren't *independent*:
 
 ```{r mvnormcor, echo=FALSE}
 xy <- mvrnorm(1000, mu = c(0, 0), Sigma = matrix(c(1, 0.8, 10, 1), nrow = 2, ncol = 2))
-plot(xy)
+plot(xy, xlab="x", ylab="y")
 ```
 
 This is useful in a mixture model, because there's no reason to think that

diff --git a/fig/rmd-02-p-fdr-1.png b/fig/rmd-02-p-fdr-1.png
diff --git a/fig/rmd-02-p-fwer-1.png b/fig/rmd-02-p-fwer-1.png
diff --git a/fig/rmd-02-plot-fdr-fwer-1.png b/fig/rmd-02-plot-fdr-fwer-1.png
diff --git a/fig/rmd-03-chooselambda-1.png b/fig/rmd-03-chooselambda-1.png
diff --git a/fig/rmd-03-coef-ridge-lm-1.png b/fig/rmd-03-coef-ridge-lm-1.png
diff --git a/fig/rmd-03-elastic-1.png b/fig/rmd-03-elastic-1.png
diff --git a/fig/rmd-03-elastic-plot-1.png b/fig/rmd-03-elastic-plot-1.png
diff --git a/fig/rmd-03-heatmap-lasso-1.png b/fig/rmd-03-heatmap-lasso-1.png
diff --git a/fig/rmd-03-plot-ridge-1.png b/fig/rmd-03-plot-ridge-1.png
diff --git a/fig/rmd-03-plotlas-1.png b/fig/rmd-03-plotlas-1.png
diff --git a/fig/rmd-05-pca-biplot-1.png b/fig/rmd-05-pca-biplot-1.png
diff --git a/fig/rmd-05-pca-biplot-ex2-1.png b/fig/rmd-05-pca-biplot-ex2-1.png
diff --git a/fig/rmd-05-pca-loadings-1.png b/fig/rmd-05-pca-loadings-1.png
diff --git a/fig/rmd-08-silhouette-1.png b/fig/rmd-08-silhouette-1.png
diff --git a/fig/rmd-10-mvnorm-1.png b/fig/rmd-10-mvnorm-1.png
diff --git a/fig/rmd-10-mvnormcor-1.png b/fig/rmd-10-mvnormcor-1.png