Hi,   I forgot to give the erreur message from R :   Error in model.frame.default(data = list(ID = c(1002L, 1003L, 1005L, 1006L,  :  variables lenght differ (found for '(weights)')   I check my variables in the dataframe and they all have the same lenght. But my outcome variable have missing values. In PROC MIXED in SAS I cas use my dataset anyway, but maybe it is not possible in R ou Zelig ?   I tried to handle missing values with an na.action=na.pass argument to pass from zelig() to lmer(), but it doesn't seem to work. The same error message comes back.   Merci, François Maurice, B. Sc., A. Stat. Candidat à la maîtrise Département de sociologie Université de Montréal --- En date de : Mer, 20.7.11, Matt <mowen(a)iq.harvard.edu> a écrit : De : Matt <mowen(a)iq.harvard.edu> Objet : Re: [zelig] Zelig seems not to pass the 'weights' argument to lmer() when using the ls.mixed model À : "Francois Maurice" <maurice.francois(a)ymail.com> Cc : "zelig(a)lists.gking.harvard.edu" <zelig(a)lists.gking.harvard.edu> Date: mercredi 20 juillet 2011 13 h 00 After briefly browsing the 'ls.mixed' model's implementation, I found nothing that immediately stuck out. Which version of Zelig are you using? Also, if you could attach a demo script on http://pastebin.com/, that might be helpful for figuring out this problem. Matt e: mowen(a)iq.harvard.edu p: 6-6132 On Jul 20, 2011, at 12:38 PM, Francois Maurice wrote: Hi,   Zelig seems not to pass the 'weights' argument to lmer() when using the ls.mixed model   The weights argument works properly when I use lmer() directly.   Is this a known bug ? Is there a fixe or a workaround ?   Merci, François Maurice, B. Sc., A. Stat. Candidat à la maîtrise Département de sociologie Université de Montréal

Hi,   I'll post a demo script soon, but for now, even lmer() doesn't work. There is this error message :   Error in asMethod(object) : matrix is not symmetric [1,2]   I mimicked the 'sleepstudy' example from lme4 package (tree variables: two numeric and one factor). To my knowledge, the only difference between my dataframe and the sleepstudy dataframe is that my outcome variable has missing values. Some subjects has not any of the 8 measures in time.   Merci, François Maurice, B. Sc., A. Stat. Candidat à la maîtrise Département de sociologie Université de Montréal --- En date de : Mer, 20.7.11, Matt <mowen(a)iq.harvard.edu> a écrit : De : Matt <mowen(a)iq.harvard.edu> Objet : Re: [zelig] Zelig seems not to pass the 'weights' argument to lmer() when using the ls.mixed model À : "Francois Maurice" <maurice.francois(a)ymail.com> Cc : "zelig(a)lists.gking.harvard.edu" <zelig(a)lists.gking.harvard.edu> Date: mercredi 20 juillet 2011 15 h 55 The error seems to indicate that the weights parameter is indeed being passed, but there is a mismatch in the length of the data.frame's rows and the weights' rows. I think we will be able to diagnose the problem with a demo script that does two things:   1. Show the code working with the 'lmer' function   2. Show the equivalent code failing with the 'zelig' function This should reveal to us the issue. Cheers - Matt e: mowen(a)iq.harvard.edu p: 6-6132 On Jul 20, 2011, at 3:30 PM, Francois Maurice wrote: Hi,   I forgot to give the erreur message from R :   Error in model.frame.default(data = list(ID = c(1002L, 1003L, 1005L, 1006L,  :  variables lenght differ (found for '(weights)')   I check my variables in the dataframe and they all have the same lenght. But my outcome variable have missing values. In PROC MIXED in SAS I cas use my dataset anyway, but maybe it is not possible in R ou Zelig ?   I tried to handle missing values with an na.action=na.pass argument to pass from zelig() to lmer(), but it doesn't seem to work. The same error message comes back.   Merci, François Maurice, B. Sc., A. Stat. Candidat à la maîtrise Département de sociologie Université de Montréal --- En date de : Mer, 20.7.11, Matt <mowen(a)iq.harvard.edu> a écrit : De : Matt <mowen(a)iq.harvard.edu> Objet : Re: [zelig] Zelig seems not to pass the 'weights' argument to lmer() when using the ls.mixed model À : "Francois Maurice" <maurice.francois(a)ymail.com> Cc : "zelig(a)lists.gking.harvard.edu" <zelig(a)lists.gking.harvard.edu> Date: mercredi 20 juillet 2011 13 h 00 After briefly browsing the 'ls.mixed' model's implementation, I found nothing that immediately stuck out. Which version of Zelig are you using? Also, if you could attach a demo script on http://pastebin.com/, that might be helpful for figuring out this problem. Matt e: mowen(a)iq.harvard.edu p: 6-6132 On Jul 20, 2011, at 12:38 PM, Francois Maurice wrote: Hi,   Zelig seems not to pass the 'weights' argument to lmer() when using the ls.mixed model   The weights argument works properly when I use lmer() directly.   Is this a known bug ? Is there a fixe or a workaround ?   Merci, François Maurice, B. Sc., A. Stat. Candidat à la maîtrise Département de sociologie Université de Montréal

Sorry, I made two mistakes in my code. I just modified it on pastebin.com.   With those corrections, the weights argument works well with lmer() for both dataframe and works well for the sleep data wiht zelig(). Within zelig(), the weights argument doesn't work with my data. Merci, François Maurice, B. Sc., A. Stat. Candidat à la maîtrise Département de sociologie Université de Montréal --- En date de : Mer, 20.7.11, Francois Maurice <maurice.francois(a)ymail.com> a écrit : De : Francois Maurice <maurice.francois(a)ymail.com> Objet : Re: [zelig] Zelig seems not to pass the 'weights' argument to lmer() when using the ls.mixed model À : "Matt" <mowen(a)iq.harvard.edu> Cc : "zelig(a)lists.gking.harvard.edu" <zelig(a)lists.gking.harvard.edu> Date: mercredi 20 juillet 2011 19 h 21 O.K., I wrote a demo script posted on pastebin.com under "Testing ls.mixed". I mimicked the sleep study from lme4 package. There is also a modified version of the sleep study data in wich I created a fake weight variable to test that data. In my knowledge, the only difference between the two dataframes is that my outcome variable has missing data.   What is weird is that the weights argument doesn't work neither in lmer() nor in zelig() and that for both dataframe.   Merci beaucoup, François Maurice, B. Sc., A. Stat. Candidat à la maîtrise Département de sociologie Université de Montréal --- En date de : Mer, 20.7.11, Matt <mowen(a)iq.harvard.edu> a écrit : De : Matt <mowen(a)iq.harvard.edu> Objet : Re: [zelig] Zelig seems not to pass the 'weights' argument to lmer() when using the ls.mixed model À : "Francois Maurice" <maurice.francois(a)ymail.com> Cc : "zelig(a)lists.gking.harvard.edu" <zelig(a)lists.gking.harvard.edu> Date: mercredi 20 juillet 2011 15 h 55 The error seems to indicate that the weights parameter is indeed being passed, but there is a mismatch in the length of the data.frame's rows and the weights' rows. I think we will be able to diagnose the problem with a demo script that does two things:   1. Show the code working with the 'lmer' function   2. Show the equivalent code failing with the 'zelig' function This should reveal to us the issue. Cheers - Matt e: mowen(a)iq.harvard.edu p: 6-6132 On Jul 20, 2011, at 3:30 PM, Francois Maurice wrote: Hi,   I forgot to give the erreur message from R :   Error in model.frame.default(data = list(ID = c(1002L, 1003L, 1005L, 1006L,  :  variables lenght differ (found for '(weights)')   I check my variables in the dataframe and they all have the same lenght. But my outcome variable have missing values. In PROC MIXED in SAS I cas use my dataset anyway, but maybe it is not possible in R ou Zelig ?   I tried to handle missing values with an na.action=na.pass argument to pass from zelig() to lmer(), but it doesn't seem to work. The same error message comes back.   Merci, François Maurice, B. Sc., A. Stat. Candidat à la maîtrise Département de sociologie Université de Montréal --- En date de : Mer, 20.7.11, Matt <mowen(a)iq.harvard.edu> a écrit : De : Matt <mowen(a)iq.harvard.edu> Objet : Re: [zelig] Zelig seems not to pass the 'weights' argument to lmer() when using the ls.mixed model À : "Francois Maurice" <maurice.francois(a)ymail.com> Cc : "zelig(a)lists.gking.harvard.edu" <zelig(a)lists.gking.harvard.edu> Date: mercredi 20 juillet 2011 13 h 00 After briefly browsing the 'ls.mixed' model's implementation, I found nothing that immediately stuck out. Which version of Zelig are you using? Also, if you could attach a demo script on http://pastebin.com/, that might be helpful for figuring out this problem. Matt e: mowen(a)iq.harvard.edu p: 6-6132 On Jul 20, 2011, at 12:38 PM, Francois Maurice wrote: Hi,   Zelig seems not to pass the 'weights' argument to lmer() when using the ls.mixed model   The weights argument works properly when I use lmer() directly.   Is this a known bug ? Is there a fixe or a workaround ?   Merci, François Maurice, B. Sc., A. Stat. Candidat à la maîtrise Département de sociologie Université de Montréal

You can just delete the cases with missing dependent variable before using zelig(). This does not change your analysis since you are assuming missing at random anyway (i.e., the observed-data likelihood will be the same). Kosuke Department of Politics Princeton University http://imai.princeton.edu On Jul 21, 2011, at 9:15 AM, Francois Maurice wrote: - Zelig Mailing List, served by Harvard-MIT Data Center Send messages: zelig(a)lists.gking.harvard.edu [un]subscribe Options: http://lists.gking.harvard.edu/?info=zelig Zelig program information: http://gking.harvard.edu/zelig/