Amelia September 2013

amelia@lists.gking.harvard.edu

5 participants
4 discussions

by Michael Gleason

Hi- It it possible to account for repeated, seasonal (i.e., monthly) spikes in Amelia? It seems that the polytime or splinetime arguments can capture moderate seasonal variability, but less adept at major spikes in, for example, heating use in December and January. So far I've tried: 1) increasing the order of splinetime to make it more flexible, 2) including month as a nominal variable, and 3) including the value from the same month in the previous and next year as additional columns in each row. All of these help, but only to a minor degree. Thanks for any ideas or guidance. Mike

10 years, 6 months

Collinearity follow up

by Alex Sutherland

Dear all, I have a couple of follow-up questions concerning the treatment of (obviously) collinear variables in Amelia II. Consider a dummy variable with missing values (y), an indicator variable indicating the missingness of y (also a dummy, say, i), and another indicator indicating the missingness of y but with reversed coding to i, say j. So whenever y is missing i=0 and j=1, when not, i=1 and j=0. A dataset with such a setting (and many other variables) has been successfully imputed with Amelia (m=21). Chain lengths vary from about 240 to 275, convergence has been reported "normal"; with empri = .03*nrow(data) (where "data" is the raw data set). From the various Amelia II discussion list responses I've seen (and responses I've got from there since asking about this), I am surprised that we did not get errors (singularity or similar). Given that we were successful with the imputation, my questions are: (1) are the results "reliable"? (2) Do we need to be concerned about the quality of the imputation? Regards, -------------------------------------------- Dr. Alex Sutherland Social Sciences' Research Methods Centre / Institute of Criminology, University of Cambridge as2140(a)cam.ac.uk http://www.crim.cam.ac.uk/people/academic_research/alex_sutherland/

10 years, 7 months

Amelia detecting & dropping collinear variables

by Alex Sutherland

Dear Amelia-users, A newbie question: Does Amelia automatically detect and ignore identical variables with different names (i.e. perfectly collinear variables) or will this cause singularity/collinearity problems ending with an error message? E.g. if one has a variable 'ethnicity' where 1==Non-white and 0=White and another variable 'White' where 1==White 0==Non-White, would one be ignored or would the imputation always fall over? From reading other messages (e.g. https://lists.gking.harvard.edu/pipermail/amelia/2012-May/000869.html) the suggestion is that one of these would need to be dropped manually or an error message would appear (as long as the variables mapped onto one another perfectly as per Matt Blackwell's response). Thanks for any assistance, and apologies if this seems ridiculously straightforward/obvious! Dr. Alex Sutherland Research Methods Associate, Social Sciences' Research Methods Centre (SSRMC) http://www.ssrmc.group.cam.ac.uk/ Office 3.12 Institute of Criminology, University of Cambridge, Sidgwick Site CB3 9DA Tel: +44 (0)1223 746519 as2140(a)cam.ac.uk http://www.crim.cam.ac.uk/people/academic_research/alex_sutherland/

10 years, 7 months

Help with Amelia: how to run Amelia in a dataset with fixed and random effects (repeated measures)

by Anup Gopalakrishnapillai

Dear Amelia community, I am a new Amelia user and I have been trying to use Amelia but could not understand how to run Amelia on a data from a study that has a within subject and a between subject variable. Below I will try to explain my situation with a hypothetical experiment and data. Experiment: Let's assume I am interested to see the reaction time of subjects to some stimulus (say they have to identify the heaviest of the objects been placed on their left and right hands) under either a neutral or a happy setting (I the later case the subjects are provided with their choice of drinks and a snack). Added to this, each subject is tested with a range of weights. And I am interested to see if the subjects with the "happy" atmosphere can perform well than the "neutral" ones. Here is the R code that makes the data for the above experiment: # ===================================================== # Create a dataframe to contain the subject name and the repeated measures factor Stimulus intensity mydata=data.frame(SubjectName=sort(rep(paste("Subject",1:8,sep="_"),8)),StimIntensity=(rep(seq(100,800,by=100)))); # Add the between subjects variable Group mydata = cbind(mydata,Group=sort(rep(c("Neutral","Happy"),32),decreasing=T)); # Creates the Response time values for first group : "Neutral" i=0;k=0;while (i < 4) {a[((i*8)+1):((i*8)+8)]=sort(rnorm(mean=10+(i*10):20+(i*10),1,n=8)); i = i+1;}; # Creates the Response time values for the second group : "Happy" i=0;b=0;while (i < 4) {b[((i*8)+1):((i*8)+8)]=sort(rnorm(mean=10+(i*10):20+(i*10),1,n=8))*c(1,1,1,1,1,1.5,1.75,2); i = i+1;}; # Add the data to the RT column in mydata dataframe mydata=cbind(mydata,"RT"=c(a,b)); # Creates a second dataframe with missing data (n=15) mydataMissing = mydata; mydataMissing[runif(n=15,min=1,max=64),"RT"] = NA; # ===================================================== As you can see I have also made an additional data fame that has the missing data. The question is how do I run Amelia in this case. Any help would be appreciated. I guess the cross section variable would be "Group" and the time series variable = "StimIntensity" but when I just provide only these it produces the error below. "amelia starting Amelia Error Code: 61 There is only 1 column of data after removing the ts, cs and idvars. Cannot impute without adding polytime." Thanks in advance Best -Anup

10 years, 7 months

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

2009

2008

2007

2006

2005

2004

2003

2002

Amelia September 2013