New package xmiss available from SSC

[Phil Clayton]

Thanks to Kit Baum, a new package xmiss is now available from SSC.

xmiss tabulates the frequency and percentage of missing values in one variable by levels of a second variable. This is useful for examining patterns of missing data.


Example:

. sysuse nlsw88
(NLSW, 1988 extract)
. xmiss race union

                                  union
                        -----------------------------
race                    Missing   Total     % missing
-----------------------------------------------------
white                   284       1637       17.3
black                   82        583        14.1
other                   2         26          7.7
-----------------------------------------------------
All                     368       2246       16.4

xmiss can be installed from SSC using:

ssc install xmiss

[Nick Cox]

For comparison, here are the same but much less compact results from missings report (http://www.statalist.org/forums/foru...ging-missings; SJ):

. bysort race: missings report union, percent

---------------------------------------------------------------------------------
-> race = white

Checking missings in union:
284 observations with missing values

union   284    17.35%

---------------------------------------------------------------------------------
-> race = black

Checking missings in union:
82 observations with missing values

union   82    14.07%

---------------------------------------------------------------------------------
-> race = other

Checking missings in union:
2 observations with missing values

union   2     7.69%

[Phil Clayton]

Thanks Nick; yes xmiss is very basic but I find the compactness (of the command as well as the results) to be handy. I should also mention that it can replace the current dataset with a report of the missing values, which I use to prepare reports of missing data patterns.