few minor updates

R/mvgam.R

@@ -884,7 +884,7 @@ mvgam = function(formula,
     # If no smooth terms are included, jagam will fail; so add a fake one and remove
     # it from the model and data structures later
     data_train$fakery <- rnorm(length(data_train$y))
-    form_fake <- update.formula(formula, ~ . + s(fakery))
+    form_fake <- update.formula(formula, ~ . + s(fakery, k = 3))
     fakery_names <- names(suppressWarnings(mgcv::gam(form_fake,
                                                      data = data_train,
                                                      family = family_to_mgcvfam(family),

R/plot_mvgam_series.R

@@ -396,13 +396,20 @@ plot_mvgam_series = function(object,
            xaxt = 'n',
            ylim = range(c(truth), na.rm = TRUE),
            xlim = c(0, length(c(truth))))
-      axis(side = 1,
-           at = floor(seq(0, max(data_train$time) -
-                            (min(data_train$time)-1),
-                          length.out = 6)),
-           labels = floor(seq(min(data_train$time),
-                              max(data_train$time),
-                              length.out = 6)))
+      if(max(data_train$time < 6)){
+        axis(side = 1,
+             at = 0:(max(data_train$time) - 1),
+             labels = 1:max(data_train$time))
+      } else {
+        axis(side = 1,
+             at = floor(seq(0, max(data_train$time) -
+                              (min(data_train$time)-1),
+                            length.out = 6)),
+             labels = floor(seq(min(data_train$time),
+                                max(data_train$time),
+                                length.out = 6)))
+      }
+
       title('Time series', line = 0)
       title(ylab = ylab, line = 1.5)
       title(xlab = "Time", line = 1.5)

R/stan_utils.R

@@ -2393,12 +2393,19 @@ if(trend_model != 'VAR1'){
            "matrix[n_series, n_lv] Z; // matrix mapping series to latent trends")
   model_file <- readLines(textConnection(model_file), n = -1)
 
+  # Z <- matrix(0, NCOL(ytimes), n_lv)
+  # for(i in 1:NROW(trend_map)){
+  #   Z[as.numeric(data_train$series)[trend_map$series[i]],
+  #            trend_map$trend[i]] <- 1
+  # }
+
   Z <- matrix(0, NCOL(ytimes), n_lv)
   for(i in 1:NROW(trend_map)){
-    Z[as.numeric(data_train$series)[trend_map$series[i]],
-             trend_map$trend[i]] <- 1
+    rowid <- which(levels(data_train$series) == trend_map$series[i])
+    Z[rowid, trend_map$trend[i]] <- 1
   }
 
+
   model_data$Z <- Z
   return(list(model_file = model_file,
               model_data = model_data))