Hello, everyone! I am currently trying to perform a three level analysis of the PISA2018 and the latest PIAAC datasets, but I am experiencing some difficulties with it. I was trying to perform this analysis in R, but as far as I have understood, at the moment it is impossible to do a multilevel analysis of complex survey designs at least for 3-levels (I think that the BIFIsurvey package allows for two-level models at the moment). That's why I turned to MLwiN, but I am quite a beginner with it so I thought I might ask for some help. The idea is to have students, nested in schools nested in countries, where I include student and school characteristics from the PISA data and then I use teachers' scores from the PIAAC data as a third level predictor. What puzzles me is this: How can I use the plausible values and the replicate weights in MLwiN and do I have to perform a separate regression with the 10 PVs and the 80 replicate weights and the final sample weight? Also, if I want to bring in to the analysis the teacher's scores, which also have PVs and replicate weights and a final sample weight, do I have to calculate it first and then add it to the PISA data by, e.g. country ID since they are the same in both datasets, or can I add it as a third level regressor externally? I would be very grateful for any comments and directions that could lead me to a successful outcome and I am looking forward to hearing your suggestions. Thank you in advance
Hi Georgi,
I recommend you to have a look at this article. It has very useful information about weighting and PV's.
Article International Large-Scale Assessment Data
For the use of PV's in MLwin, have a look at this webpage: https://www.cmm.bristol.ac.uk/forum/viewtopic.php?t=1442
Cheers!
- Badges
- Science topic
- Similar topics
- Statistics
- Survey