mmparker

No SAS, no ads, and no post-docs, please

("statistical programmer" or "statistical programming") -SAS
data scientist -marketing -advertising
applied data (analysis or analyst) -SAS -postdoc -postdoctoral -"post-doctoral"

mmparker

# This is a quick script to illustrate how to go about transforming variables
# for statistical analysis and the effects of some basic transformations.
# I'm by no means an expert on transformations, so be sure to read up on
# how to best apply the transformations!

# These two packages are for demonstrating the transformations - 
# not necessary for the transformations themselves.
library(reshape2)
library(ggplot2)

mmparker

get_random_subset <- function(df, id, n_ids = 1, select = TRUE) {

    # Pick one ID to review
    random_id <- sample(df[ , id], size = n_ids)
    
    # Print it
    df[df[ , id] %in% random_id, select]
    

mmparker

# Broadly speaking, it's safer to %in% instead of == when using a logical
# vector in R because R's indexing will return an NA - which is
# probably not the intuitive behavior. Here's what I mean:
x <- c("a", "b", NA, "c")

# Indexing with ==
x[x == "a"]

mmparker

# Calculate percent of responses by department
    
# The input here is a subset of kano_questions - all of the 
# responses related to one item
x_by_dept <- ddply(subset(x, !is.na(department)),  # Dropping that person with no department
                   .var = "department", # I'm going to split x by department
                   .fun = function(y) {   # Writing a custom, nameless function to apply
                                          # to each chunk of x
                                          

mmparker

# Dummy list
z <- list(list('a' = 1, 'b' = 2, 'c' = 3), list('a' = 4, 'b' = 5, 'c' = 6))


# If you just want to stack all of the values, not keeping any of the list name data
# but ensuring they're all of one type
data.frame(values = as.numeric(unlist(z)))

mmparker

library(zoo)


# Some sequence of Dates
x <- as.Date("2014-01-01") + c(0, 1, 2, 5, 10)


data.frame(x,
           basediff = c(NA, diff(x)),

mmparker

# Sample data
X <- data.frame(
  x = c(1, 2, 3),
  y = c(4, 5, 6),
  etc = c("a", "b", "c")
)

# Arbitrary stand-in for function that can't be vectorized (no pmax)
max.fun <- function(a, b) { max(c(a, b)) }

mmparker

@media print {

  tr{
    page-break-after: always;
    display: block;
  }
  
}

mmparker

# Calculate the date of the last day of a given quarter by pasting 
# together its first day, adding three months, and subtracting a day.
# Very elegance
# Such vectorized
calc_qtr_end <- function(year, qtr) {

    require(lubridate) # Easiest way to add a month to a date

    (as.Date(paste(year, qtr * 3, "01", sep = "-")) %m+% months(1)) - 1