help misum


misum -- Summary statistics in MI dataset


misum [varlist] [if] [in] [weight] [, m(numlist) detail format(%fmt) matrix(name)]

by is allowed

aweights, fweights and iweights are allowed. However, iweights are not allowed with detail. See summarize


misum calculates summary statistics in MI datasets. Results from summarize are combined, applying Rubin's combination rules.


+---------+ ----+ Options +----------------------------------------------------------

m(numlist) combines results from imputed datasets numlist. Numbers of imputed datasets must be given in ascending order.

detail calculates additional statistics and returns them in r(). See summarize.

format(%fmt) displays results in specified format.

matrix(name) returns the result matrix in r(name).


misum calls summarize for each imputed dataset. The returned results are added up and divided by M, where M is the number of imputed datasets. The standard deviation is calculated as the square root of the variance.

misum requires the data to be flong style (also see mi set, mi convert). The data may also be in ice format (Royston 2005).


Use with mi (Stata 11 or higher)

. sysuse auto

. mi set flong

. mi register imputed rep78

. mi impute regress rep78 = price mpg weight ,add(5)

. misum rep78 price mpg

Use with ice (Stata 9.2 or higher)

. sysuse auto ,clear

. ice rep78 price mpg weight ,clear m(5) cmd(regress)

. misum rep78 price mpg

Saved results

misum calls summarize and saves any results returned for each variable. It therefore saves the following in r():

Scalars r(varname_stat) (avarage) stat for varname returned by summarize

Matrices r(name) result matrix (matrix only)


Sebastian Pink suggested downward compatibility with Stata 9.2 and data in ice format.


Royston, P. (2005) Multiple imputation of missing values: update. Stata Journal 5 (2), 188-201.


Daniel Klein, University of Bamberg,

Also see

Online: mi, summarize

if installed: ice