############################################################
### 4. Diagnose the caliper weights ###
############################################################
caliper_weights <- function(cw_val) {
weight_df <- caliper_relevance_new(
bikes_atom,
sotw_cont,
670,
cw = cw_val,
matching_vars = matchi # matchi
)
RAL_data <- RAL_calculator(weight_df, bikes_atom)
cal_wt <- RAL_data[, .(method, weight = exp(RAL)/sum(exp(RAL))), by = .(t)]
cal_wt <- data.frame(cal_wt, cw_val)
return(data.table(cal_wt))
}
############################################################
### Actual weights for different cal widths + gewisano ###
############################################################
gg <- data.table(gen_gewisano(bikes_atom, 201, pratig = TRUE)[[2]])
colnames(gg) <- c("t", "BART", "BREGLOG", "SVBVAR")
gewisano <- data.table::melt(gg, measure.vars = c("BART", "BREGLOG", "SVBVAR"), variable.name = "method", value.name = "weight")
gewisano$cw_val <- "gewisano"
aaa <- lapply(c(0.1, 1), caliper_weights)
dfa <- do.call(rbind, aaa)
dfa <- rbind(dfa, gewisano[t>670])
ggplot(dfa, aes(x = t, y = weight, col = method)) +
geom_line() +
facet_wrap(~cw_val, ncol = 1)
ggsave("temp/cal_weights.pdf")
############################################################
### Adding the lpdenses ###
############################################################
gen_agg_preds(
bikes_atom,
start_agg = 670,
sotw_cont,
mahala = FALSE,
cw = 1
)
df_agg <- data.table(df_agg)
df_agg[t >= 670, .(meanpred = mean(lpdens)), by = method]
predabil <- RAL_data[, .(method, lpdens), by = .(t)]
ggplot(predabil, aes(x = t, y = lpdens, col = method)) + geom_line()
ggsave("temp/plot_predabil.pdf")
df_agg <- data.table(df_agg)
View(df_agg[t >= 670])
ggplot(df_agg[t >= 670], aes(x = t, y = lpdens, color = method)) + geom_line()
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.