3
$\begingroup$

I have read some posts on weighting. However, I am still unclear on the sort of weights I need to use. I am using data from the Longitudinal Survey of Australian Youth (LSAY). This survey provides longitudinal weights for each survey wave (i.e. correction for sampling error and attrition). Because I have weight variables for each period (10) and I can only specify one of them in TraMiner, I am not sure which I should use. I have read that whether weights for the first or last wave should be used. I have not read any reasons why.

Can anyone provide me some guidance on this issue?

$\endgroup$
2
$\begingroup$

TraMineR, and more generally sequence analysis, treats each sequence as a whole. Ideally, weights should therefore correct for selection bias of the sample of sequences you consider, and longitudinal weights should be used. The choice of the weights depends on which sequences you retain for your analysis:

  1. If you use only complete sequences until the last wave, then the choice should be the longitudinal weights associated to the last wave, which accounts attrition until the last wave).

  2. If you chose to retain all sequences complete until say the 6th wave and admit missing states for the last four waves, then you should chose the weights associated to wave 6.

  3. If you retain all sequences, even for those for which you have valid data for the first wave only, then you should use weights of the first wave.

  4. In case you select only sequences with no more than say 4 missing states, the choice may be more ambiguous. If most of the four missing states occur in the last four positions, then you could adopt the solution 2 as a good approximation.

Hope this helps.

$\endgroup$
1
  • 1
    $\begingroup$ Thanks Gilbert. That was exactly the kind of answer that I was looking for -clear and precise. Do you have, or know any reference that I can include in my document concerning these comments? I have spent a couple of weeks looking for one -I have not been able to find anything discussing these points in the (clear) way that you just did. $\endgroup$
    – fcorowe
    Oct 15 '13 at 6:31

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.