There are three central commands to transform variables or compute new variables from other variables:

• generate Generates new variables, egen Extensions to generate
• replace Replaces variables
• recode to recode categorical variables (details here)

General syntax

generate newvar[:lblname] =exp [if] [in]
replace oldvar =exp [if] [in]  

Where newvar is a variable to be created, oldvar an existing variable and exp an expression of any complexity using

• Variables
• Constants and system variables (e.g. _N (total number of observations in a data set), _n (number of current observations) or _pi
• Functions (help functions)
• Operators:
  • Arithmetic: +, -, *, / and ^ (raise to a power);
  • Relations == (equal), != or ~= (not equal) and >, < , >= (≥), >= (≤)
  • Logical operators & (and), | (or) and ! or ~ (not)
  • String: + (concatenation)

A large number of arithmetic functions are available; here's a partial list:

• sqrt(x), sum(x), max(x1,x2,...,xn), min(x1,x2,...,xn) , sign(x)
• Rounding and truncating functions: abs() (absolute value), ceil(x), floor(x) trunc(x) [or int(x)], mod(x,y), round(x,y) or round(x))
• Trigonometric functions like sin(), cos(), acos(),atanh() etc.
• Logarithmic functions: ln() (or log()), log10(), exp(). logit() [log of the odds ratio]
• missing() [true (1) if missing]

Several functions are shortcuts simplifying commond tasks that could be achieved with a sequence of generate/replace commands.

• Recoding functions autocode(x,n,x0,x1), recode(x,x1,x2,...,xn) (examples of use in Create groups)
• clip(x,a,b) returns x if x is in the interval a > x > b, a if x is below, b if it is above
• cond(x,a,b,c) (c can be omitted): if x is true, the result is a, if false the result is b, if it is missing the result is c.
• inrange(z,a,b) 1 (true) if z is in the interval a > z > b, 0 (false) otherwise
• Retrieving results saved by commanjds: e(name) and r(name)[value of a saved result],

This command offers a number of useful functions (some of them are documented below). The general syntax is:

egen newvar = efn(exp) [if] [in] [, options]

Where efn is one of the functions offered by egen (see a partial list below) and exp is an expression, often a simple variable name.

Functions returning a constant:

These function create a new variable with statistics obtained for each observation accross a variable list.

• The cut (recoding) function is detailed in Create groups

Consider the following two commands:

    generate nw=(v1+v2+v3+v4)/4
    egen nw=(rowmean(v1 v2 v3 v4)
 

If there are no missing values the results of the two commands will be the same; if however one or more values are missing from an observation the results will differ. In the first case a missing value will be generated for nw, in the second example an average will be computed for the non-missing values.

count(exp) counts the number of non-missing values in a variable. To obtain the number of missing values you can use the following:

egen c=count(urb)	Count the number of non-missing observation in urb
display c-_N	Display the difference between "c" and the total number of observations in the dataset (_N is a system constant)

rowmiss(varlist) and rownonmiss(varlist) can be used to inspect missing/non missing observations across several variables.

• range varname #first #last [#obs] Create a variable, e.g range xx 1 _N generates values 1,2,3 ... up to the number of observations in the current data set.

• Create groups
• Missing values
• Create dummy variables
• Reexpressions
• Documenting data (label variables)