This issue tracker has been migrated to GitHub, and is currently read-only.
For more information, see the GitHub FAQs in the Python's Developer Guide.

Author steven.daprano gregory.p.smith, ncoghlan, oscarbenjamin, remi.lapeyre, rhettinger, steven.daprano, wolma 2019-01-22.01:12:21 -1.0 Yes <20190122011141.GA13616@ando.pearwood.info>
Content
```Here is some further information on weights in statistics in general,
and SAS and Stata specifically:

https://blogs.sas.com/content/iml/2017/10/02/weight-variables-in-statistics-sas.html

Quote:

use the FREQ statement to specify integer frequencies for
repeated observations. Use the WEIGHT statement when you
want to decrease the influence that certain observations
have on the parameter estimates.

http://support.sas.com/kb/22/600.html

https://www.stata.com/manuals13/u20.pdf#u20.23

Executive summary:

- Stata defines four different kinds of weights;

- SAS defines two, WEIGHT and FREQ (frequency);

- SAS truncates FREQ values to integers, with zero or
negative meaning that the data point is to be ignored;

- Using FREQ is equivalent to repeating the data points.
In Python terms:

mean([1, 2, 3, 4], freq=[1, 0, 3, 1])

would be equivalent to mean([1, 3, 3, 3, 4]).

- Weights in SAS are implicitly normalised to sum to 1,
but some functions allow you to normalise to sum to the
number of data points, because it sometimes makes a
difference.

- It isn't clear to me what the physical meaning of weights
in SAS actually is. The documentation is unclear, it *could*
as simple as the definition of weighted mean here:

https://en.wikipedia.org/wiki/Weighted_arithmetic_mean#Mathematical_definition

but how that extends to more complex SAS functions is unclear to me.

(And for what its worth, I don't think SAS's MEAN function supports
weights at all. Any SAS users here that could comment?)```
History
Date User Action Args
2019-01-22 01:12:23steven.dapranosetrecipients: + steven.daprano, rhettinger, gregory.p.smith, ncoghlan, oscarbenjamin, wolma, remi.lapeyre