ds_count

Count the number of observations in a dataset, possibly missing or non missing for a given variable.

%ds_count(dsn, _nobs_=, miss=, nonmiss=, distinct=no, lib=WORK);

Arguments

• dsn : a dataset;
• miss : (option) the name of variable/field in the dataset for which only missing observations are considered; default: miss is not set;
• nonmiss : (option) the names of variable/field in the dataset for which only NON missing observations are considered; this is obviously compatible with the miss argument above only when the variables differ; default: nonmiss is not set;
• distinct : (option) boolean flag (yes/no) set to count only distinct values; in practice, runs a SQL SELECT DISTINCT process instead of a simple SELECT; default: no, i.e. all values are counted;
• lib : (option) name of the input library; by default: WORK is used.

Returns

_nobs_ : name of the macro variable used to store the result number of observations; by default (i.e., when neither miss nor nonmiss is set, the total number of observations is returned).

Example

Let us consider the table _dstest28:

geo	value
' '	1
AT	.
BG	2
''	3
FR	.
IT	5

then we can compute the TOTAL number of observations in _dstest28:

%local nobs;
%ds_count(_dstest28, _nobs_=nobs);

returns nobs=6, while:

%ds_count(_dstest28, _nobs_=nobs, nonmiss=value);

returns the number of observations with NON MISSING value, i.e. nobs=4, and:

%ds_count(_dstest28, _nobs_=nobs, miss=value, nonmiss=geo);

returns the number of observations with MISSING value and NON MISSING geo at the same time, i.e. nobs=1.

Run macro %_example_ds_count for more examples.

Notes

1. This macro relies on %obs_count macro since it actually runs:

%obs_count(&dsn, _ans_=&_nobs_, where=&where, pct=no, lib=&lib);

with where defined, when both miss and nonmiss parameters are passed for instance, as the SAS expression &miss is missing and not(&nonmiss is missing).

1. In practice, running the commands (which imply the creation of an intermediary table):

%ds_count(dsn, _nobs_=c0, lib=lib);
%ds_select(dsn, _tmp, where=&cond, ilib=lib);
%ds_count(_tmp, _nobs_=c1, lib=lib);

provides with a result equivalent to simply launching:

%let ans=;
%obs_count(dsn, where=&cond, _ans_=ans, pct=yes, lib=WORK);

and comparing the values of c0 and c1:

%if &c1=&c0 %then          %let ans=100;
%else %if &c1 < &c0 %then  %let ans=%sysevalf(100* &c1/&c0);
%else                      %let ans=0;

References

1. "Counting the number of missing and non-missing values for each variable in a data set".
2. Hamilton, J. (2001): "How many observations are in my dataset?".

See also

%obs_count, %var_count, %ds_check, %ds_isempty, %var_check.