timcdlucas

library(ggplot2)

Xs <- 10
N <- 200


d <- data.frame(matrix(rnorm(Xs * N), nrow = N))

d$prod <- apply(d, 1, function(x) prod(x[1:5]))

timcdlucas

library(mgcv)


# From the man page without categorical.
set.seed(2) ## simulate some data... 
dat <- gamSim(1,n=400,dist="normal",scale=2)


b <- gam(y~s(x0)+s(x1)+s(x2)+s(x3),data=dat)
summary(b)

timcdlucas

n = 100
x = runif(n)
y = 1 + 2*x + rnorm(n)

yp = predict(lm(y~x))

plot(y=yp, x=y)
abline(lm(yp~y), col="blue")
abline(a=0, b=1, col="red", lty="dashed")

timcdlucas

set.seed(10)
d <- data.frame(x = sample(c('a', 'b'), 30, replace = T),
                y = rnorm(30))

anova(lm(y~x, d))
t.test(y~x, d, var.equal = TRUE)

timcdlucas

set.seed(10)
d <- data.frame(gender = factor(sample(c('male', 'female'), 20, replace = TRUE)),
                height = factor(sample(c('tall', 'short'), 20, replace = TRUE)))

d$gender_dummy <- as.numeric(d$gender) - 1
d$height_dummy <- as.numeric(d$height) - 1
d$inter_dummy <- (as.numeric(d$gender) - 1)*(as.numeric(d$height) - 1)

timcdlucas

library(MASS)

N <- 60
x <- rep(c('a', 'b'), N)
x_dummy <- as.numeric(x == 'a')

y <- x_dummy + rt(N, 2)

data <- data.frame(x = x, y = y)

timcdlucas

library(nnet)
library(dplyr)

set.seed(20200408)

diamonds <- diamonds %>% 
              select(carat, price, depth, table, x, y, z) %>% 
              scale

timcdlucas

norm <- rep(NA, 1e4)
unif <- rep(NA, 1e4)

for(i in seq(1e4)){
  set.seed(i)
  norm[i] <- rnorm(1)
  set.seed(i)
  unif[i] <- runif(1)

timcdlucas

set.seed(17022020)

# "inside" people and countries

names_in <- c('lisa', 'suzanne', 'rohan')
countries_in <- c('germany', 'austria-hungary', 'italy', 'france')

# "outside"

names_out <- c('tim', 'jen')

timcdlucas

get_preds <- function(t){
  stopifnot(inherits(t, 'train'))
  
  oos <- t$pred %>% arrange(rowIndex)
  row_matches <- sapply(1:length(t$bestTune), function(x) oos[, names(t$bestTune)[x]] == t$bestTune[[x]])
  best_rows <- rowMeans(row_matches) == 1
  
  d <- oos[best_rows, ]