Confidence interval update.

NEWS.md

@@ -1,7 +1,7 @@
 # finalfit 1.0.61
 * bug fixes
 * fit2df.lmerMod() set default `confint_sep = " to "`. 
-* boot.predict() confidence interval limits and methods options added. 
+* boot_predict() and boot_compare() confidence interval limits and methods options added. 
 
 # finalfit 1.0.6
 * new plots example vignette https://finalfit.org/articles/all_plots_examples.html

R/boot_compare.R

@@ -6,6 +6,9 @@
 #' (see examples).
 #'
 #' @param bs.out Output from \code{boot::boot},
+#' @param confint_level The confidence level to use for the confidence interval. 
+#'   Must be strictly greater than 0 and less than 1. Defaults to 0.95, 
+#'   which corresponds to a 95 percent confidence interval.
 #' @param confint_sep String separating lower and upper confidence interval
 #' @param condense Logical. FALSE gives numeric values, usually for plotting.
 #'   TRUE gives table for final output.
@@ -23,7 +26,8 @@
 #' @examples
 #' # See boot_predict.
 
-boot_compare = function(bs.out, confint_sep = " to ", comparison = "difference", condense=TRUE,
+boot_compare = function(bs.out, confint_level = 0.95, confint_sep = " to ", 
+												comparison = "difference", condense=TRUE,
                         compare_name = NULL, digits = c(2, 3), ref_symbol = 1){
 
   if(is.null(compare_name)){
@@ -40,12 +44,15 @@ boot_compare = function(bs.out, confint_sep = " to ", comparison = "difference",
     estimate = bs_c / bs_1
     null_ref = 1
   }
-
+
+  confint_level_low = (1 - confint_level) / 2
+  confint_level_high = 1 - confint_level_low
+
   if(is.null(dim(estimate))) estimate = matrix(estimate, ncol=1) #allow single vector to pass to apply
 
   estimate_centre = apply(estimate, 2, median)
-  estimate_conf.low = apply(estimate, 2, quantile, probs = c(0.025))
-  estimate_conf.high = apply(estimate, 2, quantile, probs = c(0.975))
+  estimate_conf.low = apply(estimate, 2, quantile, probs = confint_level_low)
+  estimate_conf.high = apply(estimate, 2, quantile, probs = confint_level_high)
   estimate_p1 = apply(estimate, 2, function(x) mean(x < null_ref ))
   estimate_p2 = apply(estimate, 2, function(x) mean(x > null_ref ))
   estimate_p3 = apply(estimate, 2, function(x) mean(x == null_ref ))

R/boot_predict.R

@@ -25,7 +25,7 @@
 #' @param estimate_name Name to be given to prediction variable y-hat.
 #' @param confint_level The confidence level to use for the confidence interval. 
 #'   Must be strictly greater than 0 and less than 1. Defaults to 0.95, 
-#'   which corresponds to a 95 percent confidence interval
+#'   which corresponds to a 95 percent confidence interval.
 #' @param conf.method Passed to the type argument of boot::boot.ci(). 
 #'   Defaults to "perc". The allowed types are "perc", "basic", "bca", and "norm". 
 #'   Does not support "stud" or "all"
@@ -146,7 +146,7 @@ boot_predict = function (fit, newdata, type = "response", R = 100,
   bs.out = boot::boot(data = .data, statistic = statistic, R = R,
                         formula = formula, family = family)
 
-  bs.tidy = broom::tidy(bs.out, conf.int = TRUE, conf.level = conf.level, conf.method = conf.method)
+  bs.tidy = broom::tidy(bs.out, conf.int = TRUE, confint.level = confint.level, conf.method = conf.method)
   bs.tidy = data.frame(bs.tidy)
 
 

R/ff_merge.R

@@ -63,7 +63,7 @@ ff_merge = function(factorlist, fit2df_df, ref_symbol = "-", estimate_name=NULL,
   if(is.null(factorlist$fit_id)) stop("Include fit_id=TRUE in summary_factorlist()")
 	if(inherits(factorlist, "data.frame.ff") & "(Intercept)" %in% fit2df_df$explanatory){
 		factorlist = dplyr::bind_rows(factorlist, dplyr::tibble(label = "Intercept", fit_id = "(Intercept)")) %>% 
-			mutate(index = ifelse(is.na(index), 0, index)) %>% 
+			dplyr::mutate(index = ifelse(is.na(index), 0, index)) %>% 
 			replace(is.na(.), "-") 
 	}
   explanatory_name = names(fit2df_df)[1]

R/finalfit_internal_functions.R

@@ -825,7 +825,7 @@ globalVariables(c("L95", "U95", "fit_id", "Total", "dependent",
 									"w", "Freq", "g", "total_prop", "Prop", "index_total", "vname", "Combined",
 									"2.5 %", "97.5 %", "p.value", "estimate", "index", "n", "missing_n", "var_type",
 									"missing_percent", "var1", "var2", "keep", "label", "rowid", "term",
-									"confint_L", "confint_U", "explanatory", "p", "where"))
+									"confint_L", "confint_U", "explanatory", "p", "where", "confint.level"))
 
 
 # Workaround ::: as summary.formula not (yet) exported from Hmisc