Bryan Shalloway brshallo

# traditional `contr.sum` does not name levels, so use function from `car` package
library(car)
library(tidyverse)

param <- getOption("contrasts")
go_deviance <- param
go_deviance["unordered"] <- "contr.Sum"
options(contrasts = go_deviance)

library(tidyverse)
library(recipes)

tib_a <- tibble(x = c(rep("a", 9), "b")) %>% 
  mutate(rand = rnorm(n()))

tib_b <- tibble(x = c(rep("a", 90), "b")) %>% 
  mutate(rand = rnorm(n()))

library(dplyr)
library(tibble)
library(recipes)
library(car)

param <- getOption("contrasts")
go_deviance <- param
# traditional `contr.sum` does not name levels, so use function from `car` package
go_deviance["unordered"] <- "contr.Sum"

library(tidyverse)
library(broom)
library(car)

param <- getOption("contrasts")
go_deviance <- param
# traditional `contr.sum` does not name levels, so use function from `car` package
go_deviance["unordered"] <- "contr.Sum"

library(recipes)
library(tidyverse)

diamonds_nested <- diamonds %>% 
  group_by(cut) %>% 
  nest() %>% 
  mutate(recipes = map(data, ~recipe(price ~ clarity + color + carat, data = .x)))

diamonds_nested %&gt;%

library(forecast)
library(tidyverse)

# hyjack `forecast.stlm` and simulate rather than forecast `ets` component
# (Insufficient -- likely only partially capture variability)
devtools::source_gist("https://gist.github.com/brshallo/4b93d0cf48937da6de06ffcffaed2b57")

set.seed(1234)
ts_test <- ts(rnorm(9, 0, 3),

General approach to pseudo simulation:
Take bootstrap sample --> build model --> refit model to original data --> forecast refitted model though use simulate() on ets/trend component of forecast.

library(tidyverse)
library(forecast)

# Function takes bootstrapped sample --> fits model --> refits model to original `y`
boot_mod_refit <- function(ts, fun = stlm, ...){
  bld.mbb.bootstrap(ts, 2)[[2]] %>%

	library(tidyverse)

	# this is difficult to read, but I was proud of myself using `lobstr::ast` to figure it out so wanted to save it.
	assign_in_index <- function(start_vec, new_vals, num_index = 1:3) {
	`<-`(`[`(start_vec[num_index]), new_vals)
	start_vec
	}

	# this is the more sensible/readable version
	assign_in_index <- function(start_vec, new_vals, num_index = 1:3) {

	---
	title: "Many models"
	date: "`r paste('Last updated: ', format(Sys.time(), '%Y-%m-%d'))`"
	author: "Bryan Shalloway"
	output:
	github_document:
	toc: true
	toc_depth: 3
	html_document:
	toc: true

	simulate_stlm <- function (object, h = 2 * object$m,
	lambda = object$lambda, biasadj = NULL, newxreg = NULL,
	allow.multiplicative.trend = FALSE, ...)
	{
	if (!is.null(newxreg)) {
	if (nrow(as.matrix(newxreg)) != h) {
	stop("newxreg should have the same number of rows as the forecast horizon h")
	}
	}
	seasonal.periods <- attributes(object$stl)$seasonal.periods