[R-meta] Estimate variance from time series data
Viechtbauer, Wolfgang (SP)
wolfg@ng@viechtb@uer @ending from m@@@trichtuniver@ity@nl
Wed Aug 22 11:31:21 CEST 2018
I have heard this, or similar sentiments, before. However, if one uses an appropriate model that accounts for the dependencies among the estimates, then studies with more time steps will not automatically receive more weight. They will if the estimates are essentially independent, which is appropriate. On the other hand, if estimates are highly dependent, then this will lead to an automatic downweighting of estimates from the same study. The model in essence takes care of that for you. Also note that just looking at the weights is usually insufficient in more complex models. One really needs to look at the whole weight matrix.
From: Arne Janssen [mailto:arne.janssen using uva.nl]
Sent: Thursday, 16 August, 2018 12:57
To: Viechtbauer, Wolfgang (SP)
Cc: r-sig-meta-analysis using r-project.org
Subject: Re: [R-meta] Estimate variance from time series data
Thanks for all the suggestions, I will have a look at the suggested papers.
The reason that I did not want to use affect sizes per time step is that
studies with more time steps will then have a larger weight on the
overall analysis. Whereas I can think of some justification for this, I
would rather be on the conservative side. Meanwhile, I thought to use
the correlations of the few studies of which I do have the raw data as
an indication for the range of correlations to be expected.
Thanks again for the quick replies.
More information about the R-sig-meta-analysis