adding class 10

_posts/2023-12-11-class-9-clustering/class-9-clustering.Rmd

@@ -162,6 +162,17 @@ Repeat the process above, but compute for 4 clusters
 
 ```{r}
 # TODO find k = 4 clusters
+set.seed(0)
+km.res4 <- kmeans(x = test_data, centers = 4, nstart = 25)
+
+cluster_k4 <- cbind(test_data, "cluster" = km.res4$cluster)
+
+cluster_k4$cluster <- km.res4$cluster
+
+cluster_k4$cluster <- factor(cluster_k4$cluster)
+
+ggplot(cluster_k4, aes(x = x_val, y = y_val, color = cluster)) +
+  geom_point(size = 2)
 
 ```
 
@@ -282,6 +293,18 @@ Repeat finding clusters using different points on the tree. For example, what if
 
 ```{r}
 # TODO repeat finding clusters with different numbers of clusters
+hc_clusters5 <- cutree(tree = hc, k = 5)
+
+hc_clusters5 <- cbind(test_data, cluster = hc_clusters5)
+
+hc_clusters5$cluster <- factor(hc_clusters5$cluster)
+
+hc_clusters5 %>%
+  dplyr::mutate(sample = rownames(.)) %>%
+  ggplot(aes(x = x_val, y = y_val, label = sample, color = cluster)) +
+    geom_point(size = 2) +
+    geom_text(hjust=2, vjust=0)
+
 
 ```
 
@@ -376,7 +399,7 @@ Let's run PCA and use matrix multiplication to visualize the first PC in our x-y
 
 ```{r}
 svda <- svd(as.matrix(normalized_dat))
-pc <- as.matrix(normalized_dat) %*% svda$v[, 1] %*% t(svda$v[, 1]) # Matrix multiplication
+pv <- as.matrix(normalized_dat) %*% svda$v[, 1] %*% t(svda$v[, 1]) # Matrix multiplication
 bp <- svda$v[2, 1] / svda$v[1, 1]
 ap <- mean(pc[, 2]) - bp * mean(pc[, 1])
 dim_plot + geom_segment(xend = pc[, 1], yend = pc[, 2]) +
@@ -470,7 +493,7 @@ ggplot(pca_cluster, aes(x = PC1, y = PC2, color = cluster)) +
   ylab(paste0("PC2: ",round(percentVar[2] * 100),"% variance"))
 ```
 
-Let's try again with 3 clusters
+Let's try again with 2 clusters
 
 ```{r}
 set.seed(123)
@@ -490,7 +513,16 @@ Try with 4 clusters, what names cluster together?
 
 ```{r}
 # TODO Repeat with 4 clusters and repeat what names cluster together
+set.seed(0)
+km.res4 <- kmeans(normalized_dat, centers = 4, nstart = 25)
+
+pca_cluster <- cbind(pca_vals, "cluster" = km.res4$cluster)
 
+pca_cluster$cluster <- factor(pca_cluster$cluster)
+ggplot(pca_cluster, aes(x = PC1, y = PC2, color = cluster)) +
+  geom_point(size = 2) +
+  xlab(paste0("PC1: ",round(percentVar[1] * 100),"% variance")) +
+  ylab(paste0("PC2: ",round(percentVar[2] * 100),"% variance"))
 ```
 
 
@@ -586,17 +618,17 @@ plot(hc)
 In the plot above we see 3 clear clusters
 
 ```{r}
-hc_clusters3 <- cutree(tree = hc, k =4)
+hc_clusters3 <- cutree(tree = hc, k =3)
 
 
 hc_clusters3 <- cbind(pca_vals, "cluster" = hc_clusters3)
 
 hc_clusters3$cluster <- factor(hc_clusters3$cluster)
 
-ggplot(hc_clusters3, aes(x = PC1, y = PC3, color = cluster)) +
+ggplot(hc_clusters3, aes(x = PC1, y = PC2, color = cluster)) +
   geom_point(size = 2) +
   xlab(paste0("PC1: ",round(percentVar[1] * 100),"% variance")) +
-  ylab(paste0("PC3: ",round(percentVar[3] * 100),"% variance"))
+  ylab(paste0("PC2: ",round(percentVar[2] * 100),"% variance"))
 ```
 
 **Exercise**

_posts/2023-12-12-class-10-heatmap/class-10-heatmap.Rmd

@@ -67,7 +67,7 @@ Notice here we can only see the names for California
 
 ### Cleaning up by normalizing values
 
-It's pretty hard to see much strucutre in the data just using the raw values. Let's try with our normalized values
+It's pretty hard to see much structure in the data just using the raw values. Let's try with our normalized values
 
 ```{r, fig.height=10}
 normalized_mat <- t(t(names_mat) / colSums(names_mat))
@@ -260,7 +260,7 @@ Add your own colors to the row annotations
 We can also add annotations to the columns using the same approach. Let's load in some of the data I've compiled for the states
 
 ```{r}
-state_info <- read.csv(here("class_7-9_data", "state_info.csv"))
+state_info <- read.csv(here("class_8-10_data", "state_info.csv"))
 
 head(state_info)
 ```