Project

General

Profile

Actions
Copy link

Support #1943

open

Weights

Added by Michael Vallely over 1 year ago. Updated over 1 year ago.

Status:
Resolved
Priority:
Normal
Assignee:
Understanding Society User Support Team
Category:
Weights
Start date:
07/27/2023
% Done:

100%

Description

I am messaging with a question regarding weights. I am using data from the indresp files from waves 1 to 9 of the UKHLS. I have specified the survey design in STATA at the outset as follows:
      svyset psu [pweight = i_indinus_lw], strata(strata) singleunit(scaled)

I have two questions:
  • I am using OLS regressions for each wave of data and random effects for the pooled sample. For the OLS regressions I am using [pweight=psnenus_xd] for each wave. Is this correct?
  • I receive the message 'weights are not allowed' when I try to use weights when running random effects models. Because I have specified the survey design at the outset using i_indinus_lw as the appropriate weight, is this ok?

Thanks,
Michael

Actions
Copy link
 #1

Updated by Understanding Society User Support Team over 1 year ago

  • Status changed from New to Feedback
  • Assignee changed from Olena Kaminska to Understanding Society User Support Team
  • % Done changed from 0 to 50
  • Private changed from Yes to No

Hello Michael,

I'm about to share Pablo Cabrera's response, which I understand he replied directly to you as well.

Regarding your first question, does this mean that you are fitting 9 OLS, one for each wave (waves 1 to 9)?

Currently, you are using the design weight to fit this regression, meaning that you are controlling for the unequal probabilities of selection but ignoring household nonresponse at wave 1, individual nonresponse at wave 1 and in the following waves (attrition).

I would need more information to know which weight is adequate for your analysis, but I suggest you use the weighting FAQ document guide to select your weight (pp. 4-6).

Regarding your second point, xtreg does not accept survey weights and cannot be used with the “svy:” command. An alternative you might want to explore is to use a command from the “me” (multilevel mixed-effects model) family that allows the use of complex sampling design through “svy:” and allows to control for the fact that the observations are clustered within individuals.

I hope this information is helpful.

Best wishes,

Roberto Cavazos
Understanding Society User Support Team

Actions
Copy link
 #2

Updated by Understanding Society User Support Team over 1 year ago

  • Status changed from Feedback to Resolved
  • % Done changed from 50 to 100
Actions
Copy link

Also available in: Atom PDF