1

I'm building a multilevel model with lme4 (two levels: a class and a student level with educational data). I'm using the ICCS 2009 data: http://www.iea.nl/iccs_2009.html). ICCS 2009 advise the use weights. I made a “within-school student weight” as the product of the class and student level weight factors. I also made the sums of the weights equal to the sample size. My question: how can I specify in R that I want to use this weight as a frequency weight. Do I have to use the "survey" package?

ROLO
  • 4,183
  • 25
  • 41
Jan Modus
  • 23
  • 3
  • [How to ask a good question](http://stackoverflow.com/help/how-to-ask), and even more important: [A reproducible example](http://stackoverflow.com/questions/5963269/how-to-make-a-great-r-reproducible-example). – ROLO Sep 13 '13 at 18:12

1 Answers1

1

iccs is a complex survey (see chapter 3 of the user guide) and the R survey package cannot (yet) handle multi-level models - you can read the bottom of this post for more detail

if you don't care about standard errors, most of the functions in lme4 have a weights= argument that should give you reasonable guesses.. but the standard errors will all be wrong.

the iccs 2009 user guide recommends other software on pdf page 33 of 112.

Anthony Damico
  • 5,779
  • 7
  • 46
  • 77
  • True, but I only need to weigh at the student level. I'm not using two weights. I have the within school weight. So how can I do it? – Jan Modus Sep 15 '13 at 13:28
  • 1
    @JanModus this is a complicated setup that will need `library(survey) ; ?svrepdesign`. i'll add it to my to-do list for http://asdfree.com/ but it might not happen anytime soon.. and even then you won't be able to construct multi-level models properly. sorry :/ – Anthony Damico Sep 15 '13 at 21:03