Mixed effects random coefficients multilevel linear models SIMULATION IN R
2461
$500
 Posted:
 Proposals: 2
 Remote
 #98622
 Archived
Description
Experience Level: Expert
hello everyone. if this post looks familiar to you from a few weeks ago, you're right, it is more or less the same. i am having a very difficult time with the person i granted the project to (not from this website) and i just no longer have time to work on this by myself.
the general idea is to work on a simulation involving mixedeffects/random coefficients/multilevel linear models where one specific assumption is violated: what happens if the residuals from the fixed effects and the random effects are correlated.
i have worked out the basics of an R function by myself that generates random and fixed effects, but it doesn't quite do it the correct way. at this moment, the function takes as input:
sample size at the cluster level (level 2 if you're familair with multilevel models lingo in the social sciences like i am)
sample size within each cluster (level 1, equal number of units per cluster. no need for unbalanced situations here)
intraclasscorrelation coefficient
beta regression coefficients for the fixed effects
what i would need from the person i grant this project would be to fix the function or change it in whichever way he or she wants so that i can get what i need. wich esentially would be:
to be able to generate the residuals (both from the fixed and random effects) from the same (standard) multivariate normal distribution, so i would just need to adjust the offdiagonal elements of the covariance matrix to make them either more or less correlated
to change the function around so i can succesfully add the fixed effects (the way it is now it doesnt quite work very nicely)
if possible to be able to add as many random effects as needed. at this moment my function doesnt support the addition of random effects. if this is not possible, i think i can work somethign out with whoever gets the project.
my ideal candidate:
this has been the first time i have outsourced a project on the internet and, so far, it has not been the greatest of experiences. reviews from other websites do rank this particular one as the most likely to find reliable professionals so i am putting my business (and trust) here.
therefore, what i would ideally need is:
a) i need someone who is not only a statistician / mathematician / econometrician / quantitative methodologist but someone who is both that AND has had experience with mixedeffects/randomcoefficients/multilevel linear models because, to read the function properly, certain familiarity with them is required. i am somewhat following the notation from Harvey Goldstein (1999): "Multilevel Statistical Models" in case you need more info.
b) you can be an expert programmer in whatever software of your choice, but i *need* the finished result to be in R syntax and that can be read by my R console. i dont work or operate on SAS, Matlab, etc. and, although i know these are very powerful programs, if you submit your bid saying something like "i can do this, but i can only do this on ______" and "________" is not R or something that you can translate to R, then thankyou, but i will not be able to grant you the job.
i am sorry if i come about as somewhat harsh, but i do want to be very emphatic on the fact that i can only work with people who are reliable experts on these things. a good communication is also expected, even if it is only a short email to say "progress is being done, expect results soon" or something like that.
time is somewhat of the essence, so i'd need this in 10 days maximum.
if we work well together and are interested in repeat business for simulations in R, you can expect a looong line of simulation work to come for you. thankfully i have access to a sizeable grant and i would love to have an R expert that could help me out here in a constant basis, or to work with me as an "R coach" to guide me as to how i can finish them.
there is a paedagogical aspect to this project as well. although a working code is definitely what i'm after, i would ideally also want to understand what you're doing and why you're doing it so a wellcommented code is a must as well.
PS i will send the code to R function as soon as i receive your bid and tell me whether you think you can work with me on this or not...
thanks!
the general idea is to work on a simulation involving mixedeffects/random coefficients/multilevel linear models where one specific assumption is violated: what happens if the residuals from the fixed effects and the random effects are correlated.
i have worked out the basics of an R function by myself that generates random and fixed effects, but it doesn't quite do it the correct way. at this moment, the function takes as input:
sample size at the cluster level (level 2 if you're familair with multilevel models lingo in the social sciences like i am)
sample size within each cluster (level 1, equal number of units per cluster. no need for unbalanced situations here)
intraclasscorrelation coefficient
beta regression coefficients for the fixed effects
what i would need from the person i grant this project would be to fix the function or change it in whichever way he or she wants so that i can get what i need. wich esentially would be:
to be able to generate the residuals (both from the fixed and random effects) from the same (standard) multivariate normal distribution, so i would just need to adjust the offdiagonal elements of the covariance matrix to make them either more or less correlated
to change the function around so i can succesfully add the fixed effects (the way it is now it doesnt quite work very nicely)
if possible to be able to add as many random effects as needed. at this moment my function doesnt support the addition of random effects. if this is not possible, i think i can work somethign out with whoever gets the project.
my ideal candidate:
this has been the first time i have outsourced a project on the internet and, so far, it has not been the greatest of experiences. reviews from other websites do rank this particular one as the most likely to find reliable professionals so i am putting my business (and trust) here.
therefore, what i would ideally need is:
a) i need someone who is not only a statistician / mathematician / econometrician / quantitative methodologist but someone who is both that AND has had experience with mixedeffects/randomcoefficients/multilevel linear models because, to read the function properly, certain familiarity with them is required. i am somewhat following the notation from Harvey Goldstein (1999): "Multilevel Statistical Models" in case you need more info.
b) you can be an expert programmer in whatever software of your choice, but i *need* the finished result to be in R syntax and that can be read by my R console. i dont work or operate on SAS, Matlab, etc. and, although i know these are very powerful programs, if you submit your bid saying something like "i can do this, but i can only do this on ______" and "________" is not R or something that you can translate to R, then thankyou, but i will not be able to grant you the job.
i am sorry if i come about as somewhat harsh, but i do want to be very emphatic on the fact that i can only work with people who are reliable experts on these things. a good communication is also expected, even if it is only a short email to say "progress is being done, expect results soon" or something like that.
time is somewhat of the essence, so i'd need this in 10 days maximum.
if we work well together and are interested in repeat business for simulations in R, you can expect a looong line of simulation work to come for you. thankfully i have access to a sizeable grant and i would love to have an R expert that could help me out here in a constant basis, or to work with me as an "R coach" to guide me as to how i can finish them.
there is a paedagogical aspect to this project as well. although a working code is definitely what i'm after, i would ideally also want to understand what you're doing and why you're doing it so a wellcommented code is a must as well.
PS i will send the code to R function as soon as i receive your bid and tell me whether you think you can work with me on this or not...
thanks!
Adam K.
0% (0)Projects Completed
3
Sellers worked with
1
Jobs awarded
43%
Last project
5 Apr 2012
Canada
New Proposal
Create an account now and send a proposal now to get this job.
Sign upClarification Board Ask a Question

There are no clarification messages.