Your data analysis solution
Name is required.
Email address is required.
Invalid email address
Answer is required.
Exceeding max length of 5KB

Regression weighting not working

By Sophie Welsh | Mar 13, 2016 09:04PM CET

I'm running some ANCOVAs with each observation assigned a weight of either 0 or 1, in order to eliminate certain kinds of observations from the analysis. However, the results page gives me the same number under "removed observations" when I enter weights as when I don't. Furthermore, the number of observations used in my ANCOVAs is higher than it should be based on the number I get when I add up all the weights for each observation, so it looks like it's not eliminating enough observations. My guess is that the "removed observations" comes from the setting to "remove observations with missing data" in the ANCOVA setup, and that the weighting isn't working at all.

I've been entering the "weights" column under the tab saying "regression weights" in the ANCOVA setup window. I tried putting it under "observation weights" to see if that gave anything different, but it didn't work because that option requires the sum of weights to be at least as high as the number of observations. Do you have any advice on how I might fix this?

Thank you.

Up

0

Down

By Efthalia | Mar 14, 2016 02:37PM CET | XLSTAT Agent

Hello Sophie,

The removed observations are due to missing values. XLSTAT detects them and remove them if you activate the option “remove observations with missing data” .

The weighs you want to include should be used as observations weights and not as regression weights. The sum of weights can be lower then the number of observations so there should be any problem.
Can you please install the latest version and repeat the analysis with observations weights ?

If the problem persists, please send us your data and settings at support@xlstat.com.

Regards,
Thaila

This question has received the maximum number of answers.

Contact Us

1c26995d494fb3061dd0ae8571ffc0a4@xlstat.desk-mail.com
https://cdn.desk.com/
false
desk
Loading
seconds ago
a minute ago
minutes ago
an hour ago
hours ago
a day ago
days ago
about
false
Invalid characters found
/customer/portal/articles/autocomplete
9283