What makes a good statistical software – documentation

There are many ways to judge a statistical software, e.g. user interface, number of different statistical models it can estimate, speed, large data handling capabilities, data manipulation, complexity of scripts, graph qualities, and documentation.

This post talks about documentation. I have come crossed a number of statistical packages, Stata impressed me the most especially in terms of documentation. Its documentation is so good that you can literally use it as a text book to easily learn a new modelling routine, the history of statistics (for example how negative binomial models derives). It has many examples that can walk you through a statistical model which you are unfamiliar with, making it a pleasant learning experience.

The documentation of other packages, such as R, SAS, SPSS, I found are rather simple, with little explanation and examples, difficult to navigate through, and sometimes confusing. Recently I am exploring the weighted regression so I want to know what types of weights are available and how they differ from each other. I found this useful article online: What types of weights do SAS, Stata and SPSS support? There are four types of weights, while Stata supports all four types of weights and you would know which one to be executed by the software (actually you have to explicitly specify which weight type you want to use in your command), SPSS only supports frequency weights which is fine provided you do actually know this fact when you use “weight by” in your command (though it is not apparent from the syntax). Since SPSS only supports one type of weight, there is no confusion.

Now SAS, it is said that:

“You need to read the documentation for the proc that you are using to determine what kind of weight will be used with the weight statement.  The weight statement used in one proc might assume frequency weights while another assumes probability weights.  Some procs will handle the weights differently depending on the values of the weight variable.  For example, if all of the values of the weight variable are integers (whole numbers), SAS will assume that you have a frequency weight.  If you specify a different weight variable that has decimals, then the proc will assume that you have a probability weight.   If you cannot tell from the documentation which type of weight will be used, you will either need to do some experimenting or contact SAS technical support.”

What? You have to check the type of weight case-by-case depending on the proc (SAS procedure)? This is fine but then even so there are probabilities that you would not know how a specific proc handles your weight variable. And sometimes you have to do experiments by your own or call the SAS technical support?

I am sure SAS users will have fun with that.

This entry was posted in Subjects, Uncategorized and tagged . Bookmark the permalink.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s