Programming Stata I September 8, 2003 Eric Reinhardt (Department of Political Science, Emory University, Atlanta, GA 30322) Stata Principles
1
Stata is Stata Corp.s statistical software package. It accepts command-line input or input from a *.do file. To invoke a do file without starting Stata beforehand, just double-click on it. Always use a PC with enough RAM for the dataset you want to use. To edit a do file within Stata, type doedit [optional file name]. To run a do file within Stata, type do [file name] (use , nostop option to keep do file running regardless of errors). Starting Stata The following commands are useful for the top of a do file or an interactive session.
************************************************** * enter a short description of your program here * * author and date of program * ************************************************** cd c:\data\ /* <- type in name of working data directory here */ set mem 25m /* enter appropriately large value */ set more off /* if more is on you may have to hit the space bar a lot */ set matsize 500 /* really only matters for large-scale operations */ log using filename.log, append /* use appropriate filename */ use dataset, clear /* dataset=name of dataset you want to use */ log close /* ceases saving output to log file */ exit, STATA clear /* shuts down operations and exits Stata */
Stata Syntax NOTE: Unlike most statistical software packages, Stata programs are case-sensitive. Log Files Turn a new log file on: log using c:\working\yourlog, append. Use the , replace option instead of , append if you want to write over anything thats currently in the named logfile. Temporarily turn a running logfile off: log off. Turn it on again: log on. Cease logging and close the logfile: log close. Heres a nice trick. Want to create a log file just of the commands you are going to enter in an interactive session, skipping all the output? You could use this to create a do-file after-the-fact for what youve done (replete with any mistakes in commands, of course). Start up a command log file, which works the exact same way as the log command, but with cmdlog as the necessary command. Did you forget to start a log file but want to see what commands you have entered? Start up a new log file and then type #review x, where x is the number of previously entered command lines you want to list. Comments Any line beginning with * is ignored by the Stata processor but will appear in the log file. You can take notes during class by inserting comments like this into your class log! Any section of an input line set off inside /* */ delimiters is not executed its a comment, just like in the above do-file. Use comments to keep notes about what youre doing at this point in your program. But you cant use /* */ comments in interactive mode, just in do-files. Loading Up and Saving Datasets Open a dataset using the use command, i.e., use mydata, clear. This wipes out whatever data you currently have in memory.
Stata Programming I September 8, 2003 Eric Reinhardt (Department of Political Science, Emory University, Atlanta, GA 30322) Save a dataset like this: save mydata, replace. This wipes out whatever data was currently saved in the file of that name, if any. Just use a different name instead of mydata if you want to keep the old dataset in place! Sometimes, when dealing with huge datasets whose variables you already know the names of, you may want to open up only a subset of the datasets variables. Do this as follows: use varlist using mydata, clear. Here varlist is a list of variables separated by spaces.
2
Simple Information & Memory Management To find out what exactly is in a dataset you have loaded up, use the describe (or des) command. More extensive information can be obtained with the codebook or inspect commands. Use memory to find out how much memory is used by the dataset you have loaded up, and how much memory remains for use by Stata. If you are running into the upper bound of memory, you will need to save your work, type clear all to wipe the working memory clear, type set mem 100m (or some appropriately high number), and then load up your dataset again. You may need to assign even more memory than 100 megabytes for Statas use, depending on the size of the dataset file you want to use. (You can find out how big it is by using the Windows Explorer and view-details in the appropriate folder window.) And you effectively wont be able to use a file bigger than the amount of RAM installed in the PC you are using. Heres your excuse for applying for a grant! While you await your grant money and your new PC with 10 times as much RAM, you can still do a few things to open up and work with your very large datasets. Use the compress command to squish each variable down to the lowest memory unit that still can hold all its significant digits. Use the use varlist using mydata, clear command (see above) to open up just those variables that you need from a larger dataset. Viewing a Dataset in Detail To see your dataset in Statas spreadsheet window, type browse (and list specific variables after it if you want to see only a subset of the variables). To edit particular cells in the dataset by hand, type edit, and preserve your changes afterwards. To list the contents of a dataset into the logfile, type list, and list specific variables in the order you want them listed. You can use the if command with list. You may want to sort varlist before listing the contents, to see the values in ascending order of one or more variables. Check out help gsort for information on sorting in descending order. Abbreviations & Help Most Stata commands can be abbreviated, e.g., summarize can equally be inputted as su. To find out what the appropriate abbreviation is, take a guess, or look it up. Type help commandname, where commandname is the name of the command you want to abbreviate. The highlighted characters at the start of the command name at the top of the helpfile indicate what the abbreviation is. The helpfile on a particular command also spells out the exact syntax to be used, all the options available, a useful description of what the command actually does, and links to related commands. Use help frequently. You can also access help by clicking on the Help command on the drop-down Conditions menubar. In order to execute an operation on just a subset of the data, given that certain conditions are met, use the if option. if variable==x /* where x is another variable, expression, or particular number */ Not equals is written ~=; less than or equal to is <=; greater than or equal to is >=.
Stata Programming I September 8, 2003 Eric Reinhardt (Department of Political Science, Emory University, Atlanta, GA 30322)
3
A numeric variable is missing when it equals a period, i.e., ==.; a text variable is missing when it equals blank, i.e., ==. You may want to combine conditions, e.g., the equivalent of saying, do this if a equals a particular value and if b equals a particular value. And is typed as & and or is | (perhaps found on your backslash key). Remember, and means only if both conditions are true, further restricting the size of the subset you are working with. Or means if either condition is true, which may expand the size of the subset you are dealing with. You can execute most commands on multiple subsets of the data in one command, with the by option. Simply preface your command line with by var1: command var2... You have to sort by var1, the subset-defining variable, before doing this.
Variable Types A variable can be stored in Stata memory in a variety of different ways: see help datatypes. Of the numeric types of variables, most will be either ints (integers, i.e., no decimals) or floats (decimals allowed). floats have only 7 significant digits; if you want more, the variable must be type double. A double uses more memory than a float, which in turn uses more memory than an int. If you are not sure what you will need for a given variable, use a larger data type and then compress. To see what a variables type is, describe varname. To set the type of the variables you are about to create, set type vartype, where vartype is int, float, or double, for example. To change an existing variable to a different type, you can (1) recast newtype varlist, where newtype is a datatype name and varlist is a list of variable names to be changed; use the , force option to force the change at the cost of possibly changing the variable(s)s values accordingly; (2) replace yourvar=value, where value is an expression that would yield a non-integer result for at least one observation; or, (3) set type appropriately and then generate newvar=oldvar, where newvar is the new variable name, and oldvar is the name of the variable with the original datatype. String variables are text representations instead of simply numbers. Define the datatype as string like so: set type strx, where x is the maximum number of characters in the string. Creating a New Variable as a Function of Others To create a new variable, use the generate (or gen) command, such as gen newvar=expression. It can be used with if. If you want to change certain values of this variable, replace newvar=newexpression if.... Provide a description for your new variable by labeling it: label var newvar Description. Having defined variable type as a string, you can change a numeric variable to a string using gen newstring=string(oldvar). Change a string to a number using gen newvar=real(stringvar). Various functions exist to pick out selected pieces of a string variable (perhaps for later processing into a numeric variable): see help functions. If you want to create multiple dummy variables out of one categorical variable, one dummy for each value of the original variable, use xi: command i.origvar, for just about any command. This is useful for creating fixed effects on the fly, for example. Recoding Variables To recode particular values of a variable into new values in a single command (rather than use replace multiple times), use the recode ...