Category: R

John Mount shows us that data.table is way faster for sorting than dplyr‘s arrange function:

Notice on the above semi-log plot the run time ratio is growing roughly linearly. This makes sense: data.table uses a radix sort which has the potential to perform in near linear time (faster than the n log(n) lower bound known comparison sorting) for a range of problems (also we are only showing example sorting times, not worst-case sorting times).

In fact, if we divide the y in the above graph by log(rows) we get something approaching a constant.

John has also provided us with a markdown document for comparison.

Dave Mason continues his perusal of R data types, this time looking at the matrix:

All of the examples so far have consisted of matrices with data elements of the same class. And for good reason: it’s a requirement for a matrix. R will coerce elements with mismatched classes to the same class. Here are two vectors, one of class integer and the other of class character. After combining them into a matrix via rbind(), we see the first row of data elements are of the character class (enclosed in double quotes):

> row1 <- c(1L, 2L, 3L, 4L)
> row2 <- c("a", "b", "c", "d")
>  new_matrix <- rbind(row1, row2)
> new_matrix
     [,1] [,2] [,3] [,4]
row1 "1"  "2"  "3"  "4" 
row2 "a"  "b"  "c"  "d"

Matrices drive a large number of statistical techniques, though I tend to deal with them less directly than I would have imagined.

Sebastian Sauer shows a few methods of practical data reshaping in R:

Recoding means changing the levels of a variable, for instance changing “1” to “woman” and “2” to “man”. Binning means aggregating several variable levels to one, for instance aggregating the values From “1.00 meter” to “1.60 meter” to “small_size”.

Both operations are frequently necessary in practical data analysis. In this post, we review some methods to accomplish these two tasks.

Click through for examples of techniques you can use.

Dave Mason continues his quest to learn R, focusing on vectors.  First, he looks at vector-based mathematical operations:

Now we can determine the number of customers gained vs number of customers lost (plus/minus) for each month of the quarter by subtracting one vector from another. Each vector has the same number of elements (three), and the result is also a vector of three elements:

> net_customer_gain <- new_customers - customers_lost
> net_customer_gain
Jan Feb Mar 
-15  30   3 

The sum() function can be used to add up all the elements of a vector. Below, we get the total number of new customers and lost customers for the first quarter:

> sum(new_customers)
[1] 270
> sum(customers_lost)
[1] 252

Then he shows off subsetting in vectors:

To extract multiple elements from a vector, pass in an integer class vector to the square brackets. The values of the integer vector correspond to the elements to be extracted. Here we will extract the first, third, and fourth elements of the jersey_numbers vector:

> jersey_numbers[c(1,3,4)]
Pierce  Rondo  Allen 
    34      9     20  

The values of the integer vector can be in any order:

> jersey_numbers[c(4,1,3)]
 Allen Pierce  Rondo 
    20     34      9

Vectors are a critical part of understanding R.

John Mount has a critique of meta-packages:

Derek Jones recently discussed a possible future for the R ecosystem in “StatsModels: the first nail in R’s coffin”.

This got me thinking on the future of CRAN (which I consider vital to R, and vital in distributing our work) in the era of super-popular meta-packages. Meta-packages are convenient, but they have a profoundly negative impact on the packages they exclude.

I’m not really sold on Jones’s argument, but I do think Mount has a good point.

Ian at Fells Stats wants to make a long-running Shiny app a bit more user-friendly:

Shiny operates in a reactive programming framework. Fundamentally this means that any time any UI element that affects the result changes, so does the result. This happens automatically, with your analysis code running every time a widget is changed. In a lot of cases, this is exactly what you want and it makes Shiny programs concise and easy to make; however in the case of long running processes, this can lead to frozen UI elements and a frustrating user experience.

The easiest solution is to use an Action Button and only run the analysis code when the action button is clicked. Another important component is to provide your user with feedback as to how long the analysis is going to take. Shiny has nice built in progress indicators that allow you to do this.

There are a couple of false starts in there but by the time you reach the third act, the story makes sense.  H/T R-Bloggers