In kylebmetrum/dmutate: Mutate Data Frames with Random Variates

knitr::opts_chunk$set(comment='.')

dmutate

Mutate a data.frame, adding random variates.

library(dplyr)
library(dmutate)

Univariate examples

Some variables to use in formulae:

low_wt <- 70
high_wt <- 90
mu_wt <- 80
sd <- 60
p.female <- 0.24

Use mutate_random to implement formulae in data frame. We can put bounds on any simulated variable

data.frame(ID=1:10) %>% 
  mutate_random(WT[low_wt,high_wt] ~ rnorm(mu_wt,sd))

We can simulate from any probability distirbution in R

data.frame(ID=1:10) %>% mutate_random(X ~ rcauchy(0,0.5))

We can add the variate at any level

data.frame(ID=1:10) %>%
  mutate(GROUP = ID%%2) %>%
  mutate_random(STUDY_RE ~ rnorm(50,sqrt(50))|GROUP)

Simulate multivariate normal with bounds

mu <- c(2,200)
Sigma <- diag(c(10,1000))
XY <- X[0,] + Y[200,300] ~ rmvnorm(mu,Sigma)

The object

XY

Simulate

data.frame(ID=1:10000) %>%
  mutate_random(XY) %>% 
  summary

An extended example

data.frame(ID=1:10) %>%
  mutate(GROUP = ID%%2) %>%
  mutate_random(WT[low_wt,high_wt] ~ rnorm(mu_wt,1)) %>%
  mutate_random(STUDY_RE ~ rnorm(0,sqrt(50))|GROUP) %>%
  mutate_random(SEX ~ rbinomial(p.female)) %>%
  mutate_random(sigma ~ rgamma(1,1)) %>%
  mutate_random(kappa ~ rgamma(1,1)|GROUP) %>% signif(3)

Create formulae with expr to calculate new columns in the data.frame using dplyr::mutate

We can easily save formulae to R variables. We collect formulae together into sets called covset. For better control for where objects are found, we can specify an environment where objects can be found.

a <- X ~ rnorm(50,3)
b <- Y ~ expr(X/2 + c)
d <- A+B ~ rlmvnorm(log(c(20,80)),diag(c(0.2,0.2)))
cov1 <- covset(a,b,d)
e <- list(c=3)

Notice that b has function expr. This assigns the column named Y (in this case) to the result of evaluating the expression in the data frame using dplyr::dmutate.

.data <- data.frame(ID=1:3)

mutate_random(.data,cov1,envir=e) %>% signif(3)

kylebmetrum/dmutate documentation built on May 20, 2019, 7:30 p.m.