Evidently I didn't read your post carefully enough.
It's best to stick to Stata terminology in posing Stata questions (e.g.
in Stataspeak variables always and only means columns). There are plenty
of words Stata hasn't claimed, e.g. properties.
An orthogonal solution to my previous goes something like this:
count
local N = r(N)
local OKlist
foreach v of var rater* {
qui count if !missing(`v')
if r(N) == `N' {
local OKlist `OKlist `V'
}
}
kap `OKlist'
Here is a pseudocode translation
count the number of observations
N <- result of count
Initialise the list of OK variables to empty
cycle over variables {
count how many non-missing values in that variable
if result equals N {
add that variable to the list of OK variables
}
}
work with OKlist
This code could be specialised e.g. to apply _only_ to obs 1-49:
count in 1/49
local N = r(N)
local OKlist
foreach v of var rater* {
qui count if !missing(`v') in 1/49
if r(N) == `N' {
local OKlist `OKlist `V'
}
}
It would be possible to count missings as well, and include any variable
only if none were present.
On the useful simplicity of -count-, see
SJ-7-4 dm0033 . . . . . . Speaking Stata: Counting groups, especially
panels
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . N.
J. Cox
Q4/07 SJ 7(4):571--581 (no
commands)
discusses how to count panels through reduction commands
or through tabulation commands and how to overcome
problems that do not yield easily to these approaches
SJ-7-1 pr0029 . . . . . . . . . . . . . . . Speaking Stata: Making it
count
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . N.
J. Cox
Q1/07 SJ 7(1):117--130 (no
commands)
discusses count used with a loop over observations
or variables
Nick
[email protected]
Frank Gallo
Thank you very much for the reply and the excellent reference. Let me
start my response with I am a stata beginner. The organization of my
data table is rows (of variables) and columns (of raters). So your
syntax creates a rowmiss, which gives the count of missing values
across variables. What I want to do is exclude the raters (or columns)
that have missing values on variables (or rows) when calculating kap.
For example, I have 94 raters: rater1-rater94 columns. I have 49
variables: p1-p49 rows. If only rater1 and rater2 had missing values
on variables p1-p49, then I would want to exclude them from kap so
that I could calculate the test statistics. I do not want to drop them
from the data, though, because I have other variables such as r1-r15
for which rater1 and rater2 gave complete ratings. Simply, I want to
use the set of raters with complete ratings for the variables. I
greatly appreciate any comments. Thank you.
On Sep 4, 2009, at 1:40 PM, Nick Cox wrote:
egen rowmiss = rowmiss(rater*)
kap rater* if rowmiss == 0
See for a broader discussion
SJ-9-1 pr0046 . . . . . . . . . . . . . . . . . . . Speaking Stata:
Rowwise
(help rowsort, rowranks if installed) . . . . . . . . . . . N.
J. Cox
Q1/09 SJ 9(1):137--157
shows how to exploit functions, egen functions, and Mata
for working rowwise; rowsort and rowranks are introduced
Frank Gallo
I am running the kap command, and I want to limit the analysis to the
raters that gave complete ratings for a set of objects (or variables).
The data matrix is in the form of columns(raters) and rows(variables).
For example:
kap rater1-rater94 if rater1-rater94 != . in 1/49
I know this "does not equal missing" designation here is incorrect,
but it illustrates what I want to do and where I am stuck.
*
* For searches and help try:
* http://www.stata.com/help.cgi?search
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
