ThinkStats (by Allen B. Downey) is a good book to get you familiar with statistics (and even Python, if you’ve done some scripting in other languages).
I thought it would be interesting to present some of the examples & exercises in the book in R. Why? Well, once you’ve gone through the material in a particular chapter the “hard way”, seeing how you’d do the same thing in a language specifically designed for statistical computing should show when it’s best to use such a domain specific language and when you might want to consider a hybrid approach. I am also hoping it helps make R a bit more accessible to folks.
You’ll still need the book and should work through the Python examples to get the most out of these posts.
I’ll try to get at least one example/exercise section up a week.
Please submit all errors, omissions or optimizations in the comments section.
The star of the show is going to be the “data frame” in most of the examples (and is in this one). Unlike the Python code in the book, most of the hard work here is figuring out how to use the data frame file reader to parse the ugly fields in the CDC data file. By using some tricks, we can approximate the “field start:length” style of the Python code but still keep the automatic reading/parsing of the R code (including implicit handling of “NA” values).
The power & simplicity of using R’s inherent ability to apply a calculation across a whole column (pregnancies$agepreg <- pregnancies$agepreg / 100) should jump out. Unfortunately, not all elements of the examples in R will be as nice or straightforward.
You'll also notice that I cheat and use str() for displaying summary data.
Enough explanation! Here's the code:
# ThinkStats in R by @hrbrmstr# Example 1.2# File format info: http://www.cdc.gov/nchs/nsfg/nsfg_cycle6.htm# setup a data frame that has the field start/end infopFields <- data.frame(name = c('caseid', 'nbrnaliv', 'babysex', 'birthwgt_lb','birthwgt_oz','prglength', 'outcome', 'birthord', 'agepreg', 'finalwgt'),
begin = c(1, 22, 56, 57, 59, 275, 277, 278, 284, 423),
end = c(12, 22, 56, 58, 60, 276, 277, 279, 287, 440)
)# calculate widths so we can pass them to read.fwf()pFields$width <- pFields$end - pFields$begin + 1
# we aren't reading every field (for the book exercises)pFields$skip <- (-c(pFields$begin[-1]-pFields$end[-nrow(pFields)]-1,0))
widths <- c(t(pFields[,4:5]))
widths <- widths[widths!=0]
# read in the filepregnancies <- read.fwf("2002FemPreg.dat", widths)
# assign column namesnames(pregnancies) <- pFields$name
# divide mother's age by 100pregnancies$agepreg <- pregnancies$agepreg / 100
# convert weight at birth from lbs/oz to total ouncespregnancies$totalwgt_oz = pregnancies$birthwgt_lb * 16 + pregnancies$birthwgt_oz
rFields <- data.frame(name = c('caseid'),
begin = c(1),
end = c(12)
)rFields$width <- rFields$end - rFields$begin + 1
rFields$skip <- (-c(rFields$begin[-1]-rFields$end[-nrow(rFields)]-1,0))
widths <- c(t(rFields[,4:5]))
widths <- widths[widths!=0]
respondents <- read.fwf("2002FemResp.dat", widths)
names(respondents) <- rFields$name
str(respondents)
str(pregnancies)
