Support #814: Longitudinal weight in R with - wide datset - Understanding Society User Support

Custom queries

userforum_asset
userforum_debt
userforum_finance
userforum_wealth

Actions

Copy link

Support #814

closed

Longitudinal weight in R with - wide datset

Added by Eleonora Iob over 7 years ago. Updated over 7 years ago.

Status:

Closed

Priority:

High

Assignee:

Eleonora Iob

Category:

Weights

Start date:

07/18/2017

% Done:

100%

Description

Dear UKHLS User Support team,

I'm conducting a latent growth curve analysis using waves 2-6 in R. The dataset is in the wide format. I was wondering whether I can use the longitudinal weight provided in the last wave as my weighting variable or if any further manipulation is needed.

Kind regards,
Eleonora Iob

History
Notes
Property changes

Actions

Copy link

Updated by Peter Lynn over 7 years ago

Category set to Weights
Assignee set to Eleonora Iob
Target version set to M6
% Done changed from 0 to 50

I would have thought that the longitudinal weight should be perfectly appropriate for this case, e.g. f_indinub_lw or f_indpxub_lw or f_psnenub_lw, depending on which data sources you are using. Note the "ub"part of the variable names, which will include the BHPS samples that entered the sample at wave 2.

Actions

Copy link

Updated by Victoria Nolan over 7 years ago

Private changed from Yes to No

Actions

Copy link

Updated by Eleonora Iob over 7 years ago

Hello Peter,

Thank you for your prompt reply. I'm using f_indnsub_lw as I have data from the nurse and adult main interview.

I have a further question related to the longitudinal weight. As I'm using Maximum Likelihood Estimation, I decided to include in the analysis all participants who were present at the first time point (i.e. wave 2), although some of them dropped out at subsequent waves (N= 8856). However, f_indnsub_lw is not available for almost 2000 participants. Is this because they were not present in the last wave?

Kind regards,
Eleonora

Actions

Copy link

Updated by Peter Lynn over 7 years ago

Yes. As described in the User Guide, the longitudinal weights are defined for sample members who participated in all of a set of consecutive waves. In the case of the "ub" weights, this set is all waves from wave 2 to wave w inclusive, for the w_ version of the weight, e.g. from 2 to 6 in the case of f_indnsub_lw.

As I understand it, your analysis base is not this balanced longitudinal sample but is instead the wave 2 participants, regardless of subsequent participation behaviour. The optimum weight for you wouldtherefore be the wave 2 cross-sectional weight, b_indnsus_xw.

You might also want to add the BHPS sample participants to your analysis. They had their nurse visit at wave 3, so for them you would use c_indnsbh_xw.

Actions

Copy link