Find Jobs
Hire Freelancers

Statistical Consult: Negative Binomial Regression with Random Effects

$10-75 USD

En curso
Publicado hace casi 6 años

$10-75 USD

Pagado a la entrega
I need a statistician who can help me diagnose and fix a negative binomial regression incorporating random effects that is failing to converge. I am using R and the lme4 package, but I am open to having this work performed in STATA. The study is examining the use of parliamentary questions by Members of Parliament (MPs) from a national parliament between 1999-2016. I am using MP-level observations. The dependent variable (QW) is the number of written questions each MP asks in a particular calendar year (i.e. 2001, 2002, 2003). It is count data, so I am employing a negative binomial regression. Since I have multiple (5) observations for each MP from each of the parliamentary sessions, I am employing MP-level random effects. As a result, I have a total of 9,528 observations grouped into 1,371 groupings. The basic model I am attempting to estimate is: basemod <- ([login to view URL](WQ ~ VM + Opp + Jpart + LS_Exp + uCM + uJM + log(QDays) + (1|ID), data=qdata, verbose=T, control=glmerControl(optimizer="bobyqa",optCtrl=list(maxfun=2e5)))) WQ = the number of written parliamentary questions asked in a given year VM = the vote margin that the MP won their last election by Opp = Binary variable: 1= member of an opposition party (not in the government) Jpart = Binary variable: 1= member of a junior party in the ruling coalition (part of the government) LS_Exp = Years of parliamentary experience uCM = Years of experience as a Cabinet Minister uJM = Years of experience as a Junior Minister QDays = the number of days allotted to parliamentary questions in a given year ID = unique identifier for each MP When I attempt to run this, I get the following warnings: Warning messages: 1: In checkConv(attr(opt, "derivs"), opt$par, ctrl = control$checkConv, : Model failed to converge with max|grad| = 0.0011218 (tol = 0.001, component 1) 2: In checkConv(attr(opt, "derivs"), opt$par, ctrl = control$checkConv, : Model is nearly unidentifiable: very large eigenvalue - Rescale variables? As a result, I have tried to scale the model back to the basics (just the dependent variable and the random effects) and start building it up, but I get convergence failure even when I run a bare bones model with just the dependent variable and the random effects, I still get convergence failure. minmod <- ([login to view URL](WQ ~ (1|ID), data=qdata, verbose=T, control=glmerControl(optimizer="bobyqa",optCtrl=list(maxfun=2e5)))) Warning messages: 1: In checkConv(attr(opt, "derivs"), opt$par, ctrl = control$checkConv, : Model failed to converge with max|grad| = 0.174211 (tol = 0.001, component 1) 2: In checkConv(attr(opt, "derivs"), opt$par, ctrl = control$checkConv, : Model is nearly unidentifiable: very large eigenvalue - Rescale variables? Can you help fix this? I have access to STATA and am happy to have this work performed in STATA as well, as long as you can provide me with the code. I know that some freelancers bid on projects without reading the full description. To confirm that you have read this and understand the task, can you please make the following sentence the first line of your response: "I understand NBR using FE and I can help with this problem."
ID del proyecto: 17086839

Información sobre el proyecto

4 propuestas
Proyecto remoto
Activo hace 6 años

¿Buscas ganar dinero?

Beneficios de presentar ofertas en Freelancer

Fija tu plazo y presupuesto
Cobra por tu trabajo
Describe tu propuesta
Es gratis registrarse y presentar ofertas en los trabajos
Adjudicado a:
Avatar del usuario
I understand NBR with FE I am familiar with NBR from analysing RNAseq data. Mixed effects models with individual level random effects can be recast as GEE models with a single parameter for the correlations between individual level scores. The GEE model estimates the parameters averaged over the whole population, with the correlation treated as a nuisance parameter. It therefore functions well in balanced datasets with large numbers of individuals and small numbers of repeat measurements, as would seem to be the case here. The so called sandwich estimator is robust to misspecification of the correlation model in large samples. It is also computationally robust and likely to converge even when the mixed model does not. I am a professional statistical consultant. Please see the resume on my profile for credentials. Faithfully, Tom Price
$83 USD en 1 día
0,0 (1 comentario)
0,0
0,0
4 freelancers están ofertando un promedio de $101 USD por este trabajo
Avatar del usuario
I'm an Engineer in Statistics and Applied Economics, i've got many skills and expertise that will allow me achieving perfectly all the projects and assignments because my priority is your satisfaction I am a master of softwares like SPSS, SAS, STATA, R, Eviews, Minitab, MS excel I have advanced knowledges in: Econometrics: Panel Data, Pooled Data, cross-section Data, Probit, Logit, Multinomial Logit, scoring, Economics modeling, Tobit, Heckman.. ---Data Mining: Factor analysis, Principal Component Analysis, Regression (Simple, Multiple, logisitic, hierarchical, Poisson), PCA, PCA, Data Quality, Data Reliability, ANOVA, ANOM, ANCOVA, Clustering, Hierarchical Clustering and bayesian analysis I'll provide you perfect reporting using MsWord. At the same time i can help you to do any statistics related Reports Presentations Homework Tutorials Questionnaires I have Three years of research experience. I will provide the best quality work available within minimum time
$100 USD en 7 días
4,9 (150 comentarios)
6,5
6,5
Avatar del usuario
"I understand NBR using FE and I can help with this problem." Hi, I am interested to complete the assignment using STATA. I have expertise in panel data modelling using STATA. Please visit my public profile. I can check the model setting in STATA and can fix the problem if it actually fixable. One of the reason for this problem may be due to large number of panels but I am not sure this time. Please feel free to discuss further. Thanks.
$80 USD en 3 días
4,6 (80 comentarios)
6,3
6,3
Avatar del usuario
Hello, I can help with you in your project Statistical Consult: Negative Binomial Regression with Random Effects. I have more than 5 years of experience in Mathematics, R Programming Language, Statistical Analysis, Statistics. We have worked on several similar projects before! We have worked on 300+ Projects. Please check the profile reviews. I can deliver your job with in your deadline. Please ping me for more discussion. I can assure the 100% job satisfaction. Thanks,
$140 USD en 3 días
4,8 (30 comentarios)
5,8
5,8

Sobre este cliente

Bandera de UNITED KINGDOM
London, United Kingdom
5,0
35
Forma de pago verificada
Miembro desde jul 7, 2016

Verificación del cliente

¡Gracias! Te hemos enviado un enlace para reclamar tu crédito gratuito.
Algo salió mal al enviar tu correo electrónico. Por favor, intenta de nuevo.
Usuarios registrados Total de empleos publicados
Freelancer ® is a registered Trademark of Freelancer Technology Pty Limited (ACN 142 189 759)
Copyright © 2024 Freelancer Technology Pty Limited (ACN 142 189 759)
Cargando visualización previa
Permiso concedido para Geolocalización.
Tu sesión de acceso ha expirado y has sido desconectado. Por favor, inica sesión nuevamente.