Apply generate to a matrix
mgen exprlist , in(matname) out(matname) [ common(term) ]
where exprlist is
[[`]"]expr1["[']] [[`]"]expr2["[']] [[`]"]expr3["[']] ...
and a single expr is
(important notes: (1) expr must not contain spaces unless enclosed in quotes; (2) the exprs are processed one after another, i.e., expr2 can refer to the column created by expr1, but not the other way round; (3) be careful with replacing columns and then referring to the replaced column in a later expr)
mgen interprets the columns of a matrix as variables, generates new columns (or replaces old columns) according to a specified list of expressions, and then returns the generated (and the replaced) columns as a new matrix.
Note: The column names of the input matrix and the specified new column names should conform to the naming conventions for variables.
in(matname) specifies the input matrix.
out(matname) specifies the output matrix.
common(term) specifies a common term which is added to the end of each of the expressions.
. mat A=(1,2\3,4) . mat coln A=var1 var2 . mat list A A[2,2] var1 var2 r1 1 2 r2 3 4
. mgen var3=var1+var2 var4=var3^2, in(A) out(B) B[2,2] var3 var4 r1 3 9 r2 7 49
Advanced application (note: tabstatmat is available from SSC):
. tabstat e ft, by(year) s(mean) save not . tabstatmat A . mgen ne=(1-e) pt=e*(1-ft) ft=e*ft total=(ne+pt+ft)/100 , in(A) out(B) common(*100)
Assume, "e" is labor force status (1 = in labor force, 0 = not in labor force) and "ft" is employment type (1 = fulltime, 0 = parttime, . = missing/not in labor force). The rows of matrix B will then contain the percentage distribution of nonemployed, partime employed and fulltime employed, e.g.:
B[4,4] ne pt ft total 1970 52.314139 5.8175698 41.868291 100 1980 51.818218 6.9509233 41.230859 100 1990 47.904043 9.7894623 42.306494 100 2000 48.004791 13.477444 38.517765 100
Ben Jann, ETH Zurich, email@example.com
Manual: [U] 14.3 Naming conventions