Fix constructionModelesLassoMLE: phiInit was reshaped from array to matrix. Add examp...
[valse.git] / pkg / R / plot_valse.R
1 #' Plot
2 #'
3 #' It is a function which plots relevant parameters
4 #'
5 #' @param X matrix of covariates (of size n*p)
6 #' @param Y matrix of responses (of size n*m)
7 #' @param model the model constructed by valse procedure
8 #' @param comp TRUE to enable pairwise clusters comparison
9 #' @param k1 index of the first cluster to be compared
10 #' @param k2 index of the second cluster to be compared
11 #'
12 #' @importFrom ggplot2 ggplot aes ggtitle geom_tile geom_line geom_point scale_fill_gradient2 geom_boxplot theme
13 #' @importFrom cowplot background_grid
14 #' @importFrom reshape2 melt
15 #'
16 #' @export
17 plot_valse <- function(X, Y, model, comp = FALSE, k1 = NA, k2 = NA)
18 {
19 n <- nrow(X)
20 K <- length(model$pi)
21 ## regression matrices
22 gReg <- list()
23 for (r in 1:K)
24 {
25 Melt <- melt(t((model$phi[, , r])))
26 gReg[[r]] <- ggplot(data = Melt, aes(x = "Var1", y = "Var2", fill = "value")) +
27 geom_tile() + scale_fill_gradient2(low = "blue", high = "red", mid = "white",
28 midpoint = 0, space = "Lab") + ggtitle(paste("Regression matrices in cluster", r))
29 }
30 print(gReg)
31
32 ## Differences between two clusters
33 if (comp)
34 {
35 if (is.na(k1) || is.na(k2))
36 print("k1 and k2 must be integers, representing the clusters you want to compare")
37 Melt <- melt(t(model$phi[, , k1] - model$phi[, , k2]))
38 gDiff <- ggplot(data = Melt, aes(x = "Var1", y = "Var2", fill = "value"))
39 + geom_tile()
40 + scale_fill_gradient2(low = "blue", high = "red", mid = "white", midpoint = 0,
41 space = "Lab")
42 + ggtitle(paste("Difference between regression matrices in cluster",
43 k1, "and", k2))
44 print(gDiff)
45 }
46
47 ### Covariance matrices
48 matCov <- matrix(NA, nrow = dim(model$rho[, , 1])[1], ncol = K)
49 for (r in 1:K)
50 matCov[, r] <- diag(model$rho[, , r])
51 MeltCov <- melt(matCov)
52 gCov <- ggplot(data = MeltCov, aes(x = "Var1", y = "Var2", fill = "value")) + geom_tile()
53 + scale_fill_gradient2(low = "blue", high = "red", mid = "white", midpoint = 0,
54 space = "Lab")
55 + ggtitle("Covariance matrices")
56 print(gCov)
57
58 ### Proportions
59 gam2 <- matrix(NA, ncol = K, nrow = n)
60 for (i in 1:n)
61 gam2[i, ] <- c(model$proba[i, model$affec[i]], model$affec[i])
62
63 bp <- ggplot(data.frame(gam2), aes(x = "X2", y = "X1", color = "X2", group = "X2"))
64 + geom_boxplot()
65 + theme(legend.position = "none")
66 + background_grid(major = "xy", minor = "none")
67 print(bp)
68
69 ### Mean in each cluster
70 XY <- cbind(X, Y)
71 XY_class <- list()
72 meanPerClass <- matrix(0, ncol = K, nrow = dim(XY)[2])
73 for (r in 1:K)
74 {
75 XY_class[[r]] <- XY[model$affec == r, ]
76 if (sum(model$affec == r) == 1) {
77 meanPerClass[, r] <- XY_class[[r]]
78 } else {
79 meanPerClass[, r] <- apply(XY_class[[r]], 2, mean)
80 }
81 }
82 data <- data.frame(mean = as.vector(meanPerClass),
83 cluster = as.character(rep(1:K, each = dim(XY)[2])), time = rep(1:dim(XY)[2], K))
84 g <- ggplot(data, aes(x = "time", y = "mean", group = "cluster", color = "cluster"))
85 print(g + geom_line(aes(linetype = "cluster", color = "cluster"))
86 + geom_point(aes(color = "cluster")) + ggtitle("Mean per cluster"))
87 }