We Need You!

[Michael Betancourt]

Stanis,

In preparation for some upcoming Stan talks we're hoping to accumulate a list of some of the projects to which people have been applying Stan. If you're using Stan for an existing project or even experimenting with such an implementation, we'd love for you to pass along your field so that we can get an idea of the breadth of our users. Of course, more information is always welcome.

Thanks!

-Mike Betancourt for the Development Team

[Guy Freeman]

I'm happily using Stan as part of my research into influenza epidemiology. I want to reference it correctly in my soon-to-be-submitted manuscript, so could you confirm the best way for me to do so?

Thanks for this wonderful project,

Guy

[Michael Betancourt]

Guy,

Please see http://mc-stan.org/citations.html. Thanks!

> --
> You received this message because you are subscribed to the Google Groups "stan users mailing list" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to stan-users+...@googlegroups.com.
> For more options, visit https://groups.google.com/groups/opt_out.
>
>

[Jeffrey Arnold]

Political science / international relations. Using Stan for a dynamic factor model of bond prices with the factors modeled as regressions with war events as covariates.  It's still a work in progress, so I don't have a draft yet.

[Shige Song]

Demography/sociology. I am experimenting with Stan and plan to use it in my study of infant mortality and sex ratio variations.

Shige

On Mon, Feb 4, 2013 at 6:02 PM, Michael Betancourt <betan...@gmail.com> wrote:

[W Robert Long]

Cardiovascular and substance-abuse epidemiology.

So far I haven't been successful in getting Stan to work with my data,
but that is likely a reflection of my inexperience with Stan.

I have interest in using Stan to fit a zero-inflated negative-binomial
mixed effects model, with random intercepts. This is a repeated measures
study (10 measurement occasions, 100 subjects). with attrition where I
would also like to model the missing data. The outcome is zero inflated
and I previously used a ZINB model (fitted with glmmADMB) with a smaller
subset of the data (shorter study duration), although there were
convergence problems with glmmADMB, particularly on some imputed
datasets (I previously used multiple imputation). Does this sound like a
model that could be fit with Stan ?

[Simon Blomberg]

Evolutionary biology. I am experimenting with using Stan to fit models with correlated residuals due to shared evolutionary history among species.

Cheers,

Simon.

On Mon, Feb 4, 2013 at 6:02 PM, Michael Betancourt <betan...@gmail.com> wrote:

Stanis,

In preparation for some upcoming Stan talks we're hoping to accumulate a list of some of the projects to which people have been applying Stan.  If you're using Stan for an existing project or even experimenting with such an implementation, we'd love for you to pass along your field so that we can get an idea of the breadth of our users.  Of course, more information is always welcome.

Thanks!

-Mike Betancourt for the Development Team

--
You received this message because you are subscribed to the Google Groups "stan users mailing list" group.
To unsubscribe from this group and stop receiving emails from it, send an email to stan-users+...@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.
 
 

-- 
Simon Blomberg, BSc (Hons), PhD, MAppStat, AStat.
Lecturer and Consultant Statistician
School of Biological Sciences
The University of Queensland
St. Lucia Queensland 4072
Australia
T: +61 7 3365 2506
email: S.Blomberg1_at_uq.edu.au
http://www.evolutionarystatistics.org

Policies:
1.  I will NOT analyse your data for you.
2.  Your deadline is your problem.

Statistics is the grammar of science - Karl Pearson.

[Matt]

I'm currently experimenting with it for multivariate longitudinal neuropsychology data, and I've also had a little experiment with it for hierarchical non-linear fitting for a neuropsychopharmacology/psychophysiology project. But nothing to completion with stan yet.

Matt Albrecht

On Tuesday, 5 February 2013 07:02:47 UTC+8, Michael Betancourt wrote:

[Pablo E. Verde]

Dear Stanautas,

I am working with Stan in generalized evidence synthesis (complex meta-analysis models), my work has been partially finance by a grant from the German Research Fundation (DFG Oh 39/11-1). The areas of medical applications are psiquiatry and diabetology.

Cheers,

Pablo




El 06/02/13 00:33, Simon Blomberg escribió:

[Zhenming Su]

I am using Stan as the estimation engine testing the performance of a Bayesian state-space model for fish population dynamics modeling.

Zhenming

[Richard McElreath]

I am an evolutionary anthropologist. I have two manuscripts in the wild that used Stan for modeling.

(1) An analysis of serotonin receptor variation and behavior, using Stan for some ordered logit models. Nothing fancy, just nice to have solid convergence with arbitrary varying effect structures. Also, ordered logit in Stan has been much faster for me than doing it in JAGS, I assume because of the awkward intermediate nodes needed in a JAGS implementation (?). This paper is in press at PNAS, so embargoed, but I could forward the citation along, once I have page numbers.

(2) A paper on human foraging returns across the lifespan, using multilevel gamma regressions with imputation. Most complex models have tens-of-thousands of parameters. Stan has saved my bacon, here. This paper is under review.

On the experimental end, I've gotten Stan to do a nice job on dynamic social learning data. The models are variants of "experience weighted attraction" models, with varying effects added. These models work much nicer in Stan than in other engines I've tried, I think because all of the calculations that go into computing the "attraction" scores get compiled down to lean C++ code in Stan, so aren't interpreted during sampling.

I also wrote a convenience R package, glmer2stan, to generate basic GLMM Stan models from lme4-style model formulas. This has been useful in teaching, mainly. I've been meaning to migrate it's repository over to github. So once Rstan appears there, I'll see about linking it up.

[Sebastian Weber]

Hi!

I used Stan in Systems Biology to fit longitudinal data from growing cell populations. By now I switched field and now work in pharmacometrics where I already applied it to extract from titration data parameters of a four parameter logistic function needed to determine IC50 values. Many of my Bayesian colleagues are very interested...

Best,

Sebastian

> --
> You received this message because you are subscribed to the Google Groups "stan users mailing list" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to stan-users+...@googlegroups.com.
> For more options, visit https://groups.google.com/groups/opt_out.
>
>

--

Sebastian Weber

[Phil Price]

I'm using Stan to model electricity use vs time in commercial buildings. There are several applications. My main interests at the moment are:

1. Creating a "baseline model" to predict how much electricity a building will use over a short time interval (such as 4 to 6 hours) under business as usual. An intervention is then performed that lasts a few hours, and the question is, how much energy did the intervention save? We subtract the actual energy from the predicted energy to get an estimate of this. My models use time-of-day and time-of-week indicator variables, plus a non-linear temperature response, plus an autoregressive noise term. All parameters can change from one week to the next, but there are constraints about the changes (based on prior knowledge of how building schedules work). 

2. Detecting faults and anomalies. Week-to-week changes in parameters (such as the relationship between outdoor air temperature and electricity consumption, at a given time of day) should usually be small; a big increase usually means something has gone wrong.

The usual approaches to these problems are quite simplified. With Stan I can fit a model that is much closer to describing the way a building really works (for instance, some schedule changes affect every weekday; others affect only certain days of the week; and if a change affects one 15-minute interval then it probably affects the subsequent one as well; and so on). 

My biggest problem is speed: it takes hours for my model to run on a realistically useful amount of data (say, 10 weeks of data collected every 15 minutes), which makes extending it and testing it a very slow process. But it's much much better than any alternative I know of. 

Phil Price

Lawrence Berkeley National Laboratory

[Christian Roy]

I am using Stan to model ducks population dynamic. I am interested in quantifying (and hopefully explaining) the correlation between populations that have been surveyed annually since the last 50 years.  

I am also using stan for a species community analysis.  

Chris

-----

Christian Roy, Biol. Msc.
Étudiant au Doctorat / Phd Student
Université Laval
Faculté de foresterie, de géographie et de géomatique
2405, rue de la Terrasse, bureau 2107
Québec (QC), G1V 0A6
Tél: (418) 656-2131 ext 6110

[Bob Carpenter]

Yes, that sounds like you could fit it with Stan.  Do you have a BUGS/JAGS model to start from?  Or a mathematical definition of the full probability function?

Because of Stan's imperative nature, it's easy to fit things like zero-inflated Poisson models.  I just revised our appendix to match the definition in (Lambert 1992) of zero-inflated Poisson distributions.

I've attached a PDF of the revised manual page.

- Bob