When you implement a statistical algorithm in a vector-matrix language such as SAS/IML, R, or MATLAB, you should measure the performance of your implementation, which means that you should time how long a program takes to analyze data of varying sizes and characteristics. There are some general tips that can

## Tag: **Efficiency**

As my colleague Margaret Crevar recently wrote, it is useful to know how long SAS programs take to run. Margaret and others have written about how to use the SAS FULLSTIMER option to monitor the performance of the SAS system. In fact, SAS distributes a macro that enables you to

Imagine that you have one million rows of numerical data and you want to determine if a particular "target" value occurs. How might you find where the value occurs? For univariate data, this is an easy problem. In the SAS DATA step you can use a WHERE clause or a

Evaluating a cumulative distribution function (CDF) can be an expensive operation. Each time you evaluate the CDF for a continuous probability distribution, the software has to perform a numerical integration. (Recall that the CDF at a point x is the integral under the probability density function (PDF) where x is

Friends have to look out for each other. Sometimes this can be slightly embarrassing. At lunch you might need to tell a friend that he has some tomato sauce on his chin. Or that she has a little spinach stuck between her teeth. Or you might need to tell your

A common task in SAS/IML programming is finding elements of a SAS/IML matrix that satisfy a logical expression. For example, you might need to know which matrix elements are missing, are negative, or are divisible by 2. In the DATA step, you can use the WHERE clause to subset data.

A SAS customer showed me a SAS/IML program that he had obtained from a book. The program was taking a long time to run on his data, which was somewhat large. He was wondering if I could identify any inefficiencies in the program. The first thing I did was to

My last blog post showed how to simulate data for a logistic regression model with two continuous variables. To keep the discussion simple, I simulated a single sample with N observations. However, to obtain the sampling distribution of statistics, you need to generate many samples from the same logistic model.

In my book Simulating Data with SAS, I show how to use the SAS DATA step to simulate data from a logistic regression model. Recently there have been discussions on the SAS/IML Support Community about simulating logistic data by using the SAS/IML language. This article describes how to efficiently simulate

In a previous post, I stated that you should avoid matrix multiplication that involves a huge diagonal matrix because that operation can be carried out more efficiently. Here's another tip that sometimes improves the efficiency of matrix multiplication: use parentheses to prevent the creation of large matrices. Matrix multiplication is

I love working with SAS Technical Support because I get to see real problems that SAS customers face as they use SAS/IML software. The other day I advised a customer how to improve the efficiency of a computation that involved multiplying large matrices. In this article I describe an important

SAS/IML programmers know that the VECDIAG matrix can be used to extract the diagonal elements of a matrix. For example, the following statements extract the diagonal of a 3 x 3 matrix: proc iml; m = {1 2 3, 4 5 6, 7 8 9}; v = vecdiag(m); /* v = {1,5,9}

How do you count the number of unique rows in a matrix? The simplest algorithm is to sort the data and then iterate down the rows, comparing each row with the previous row. However, this algorithm has two shortcomings: it physically sorts the data (which means that the original locations

The power of a statistical test measures the test's ability to detect a specific alternate hypothesis. For example, educational researchers might want to compare the mean scores of boys and girls on a standardized test. They plan to use the well-known two-sample t test. The null hypothesis is that the

Last week someone posted an interesting question to the SAS/IML Support Community. The problem involved four nested DO loops and took hours to run. By transforming several nested DO loops into an equivalent matrix operation, I was able to reduce the run time to about one second. The process of

The SAS/IML language secretly creates temporary variables. Most of the time programmers aren't even aware that the language does this. However, there is one situation where if you don't think carefully about temporary variables, your program will silently produce an error. And as every programmer knows, silent wrong numbers are

A while ago I saw a blog post on how to simulate Bernoulli outcomes when the probability of generating a 1 (success) varies from observation to observation. I've done this often in SAS, both in the DATA step and in the SAS/IML language. For example, when simulating data that satisfied

In a recent article on efficient simulation from a truncated distribution, I wrote some SAS/IML code that used the LOC function to find and exclude observations that satisfy some criterion. Some readers came up with an alternative algorithm that uses the REMOVE function instead of subscripts. I remarked in a

Last week I wrote about using acceptance-rejection algorithms in vector languages to simulate data. The main point I made is that in a vector language it is efficient to generate many more variates than are needed, with the knowledge that a certain proportion will be rejected. In last week's article,

A few days ago on the SAS/IML Support Community, there was an interesting discussion about how to simulate data from a truncated Poisson distribution. The SAS/IML user wanted to generate values from a Poisson distribution, but discard any zeros that are generated. This kind of simulation is known as an

The other day I was constructing covariance matrices for simulating data for a mixed model with repeated measurements. I was using the SAS/IML BLOCK function to build up the "R-side" covariance matrix from smaller blocks. The matrix I was constructing was block-diagonal and looked like this: The matrix represents a

Last week I wrote an article in which I pointed out that many SAS programmers write a simulation in SAS by writing a macro loop. This approach is extremely inefficient, so I presented a more efficient technique. Not only is the macro loop approach slow, but there are other undesirable

Over the past few years, and especially since I posted my article on eight tips to make your simulation run faster, I have received many emails (often with attached SAS programs) from SAS users who ask for advice about how to speed up their simulation code. For this reason, I

I have blogged about three different SAS/IML techniques that iterate over categories and process the observations in each category. The three techniques are as follows: Use a WHERE clause on the READ statement to read only the observations in the ith category. This is described in the article "BY-group processing

"Help! My simulation is taking too long to run! How can I make it go faster?" I frequently talk with statistical programmers who claim that their "simulations are too slow" (by which they mean, "they take too long"). They suspect that their program is inefficient, but they aren't sure why.

In a previous blog, I showed how to use SAS/IML subscript reduction operators to compute the location of the maximum values for each row of a matrix. The subscript reduction operators are useful for computing simple statistics for each row (or column) of a numerical matrix. If x is a

The birthday matching problem is a classic problem in probability theory. The part of it that people tend to remember is that in a room of 23 people, there is greater than 50% chance that two people in the room share a birthday. But the birthday matching problem is also

Sometimes in matrix computations you need to obtain the values of certain submatrices such as the diagonal elements or the super- or subdiagonal elements. About a year ago, I showed one way to do that: convert subscripts to indices and vice-versa. However, a tip from @RLangTip on Twitter got me

I recently blogged about Mahalanobis distance and what it means geometrically. I also previously showed how Mahalanobis distance can be used to compute outliers in multivariate data. But how do you compute Mahalanobis distance in SAS? Computing Mahalanobis distance with built-in SAS procedures and functions There are several ways to

Way back when I learned to program, I remember a computer instructor explaining that an IF-THEN statement can be a relatively slow operation. He said "If a multiplication takes one unit of time, an IF statement requires about 70 units." I don't know where his numbers came from, or even