user4733

590 reputation

111

bio	website
	location
	age
visits	member for	1 year, 11 months
	seen	May 13 at 17:58
stats	profile views	43

	590 reputation	bio	website		visits	member for	1 year, 11 months
111 badges		location			seen	May 13 at 17:58

summary answers questions tags badges favorites bounties reputation activity

62 Actions

suggestions reviews revisions comments badges posts accepts all

Apr 22	awarded	Tumbleweed
Apr 16	comment	Can any probability distribution be written as a Boltzmann distribution? I may have confused the issue by bringing up implementation, as that may vary widely depending on the problem. When I mentioned "special case" - I was implying that a true "infinity" could be hard coded in some simulations. This can be problematic, but maybe it's better to regard implementation issues separately.
Apr 16	comment	Can any probability distribution be written as a Boltzmann distribution? I don't understand the criticism regarding the integral not being finite - the partition function is defined such that the resulting probability integrates to 1. E(x)=infinity regions of the space simply wouldn't contribute to the probability integral.
Apr 16	comment	Can any probability distribution be written as a Boltzmann distribution? That's a good point. It's commonplace to assign values of infinity to regions of the energy functions - e.g. hard sphere potentials, square-well approximations to LJ potentials, self-avoiding walk models of polymers, lattice models of proteins etc. In practical implementations of simulations or samplers, it could be implemented as either a "large-enough" number or as a special case.
Apr 16	answered	Can any probability distribution be written as a Boltzmann distribution?
Apr 15	revised	Changing the effect measure flips the estimated direction in a Bayesian difference-in-difference analysis deleted 46 characters in body
Apr 15	asked	Changing the effect measure flips the estimated direction in a Bayesian difference-in-difference analysis
Apr 5	comment	Accuracy Measurement in Biased Dataset You must implicitly have another definition of accuracy in mind. You'll need toa articulate what that is to have an answer.
Apr 5	comment	Mixture model as a prior distribution If what you're asking is whether hierarchical models can be used to model mixture distributions and infer latent mixture components, the answer is yes.
Apr 4	comment	Mixture model as a prior distribution I don't understand this question. A hierarchical model often implies that a posterior predictive distribution on a parameter is a mixture distribution, so maybe there's some confusion here. For a particular parameter, the prior within the hierarchy can be any kind of distribution. However, it's more common to use something simple, like a conjugate prior and allow the complexity/mixture modeling to come from the hierarchical structure itself.
Apr 4	revised	How to make sure explaining variable actually explains explained variable, and not vice versa? edited body
Apr 4	answered	Is there a way in python/java/scala to convert/normalize log normal distribution into normal distribution?
Apr 4	revised	How to make sure explaining variable actually explains explained variable, and not vice versa? added 158 characters in body
Apr 4	revised	How to make sure explaining variable actually explains explained variable, and not vice versa? added 43 characters in body
Apr 4	answered	How to make sure explaining variable actually explains explained variable, and not vice versa?
Oct 10	comment	Reporting contrasts between binary probability parameters in Bayesian data analysis - odds ratios or difference in probability? In my case, there are several comparisons on different scales, but I'd like to present results in a uniform way. My impression is posteriors are often presented on a 0-1 scale (makes sense since the model is on that scale), but it's awkward to present the results as such when most of the literature is based on logistic regression and presents odds ratios. Is there not a consensus on how to present such results in Bayesian data analyses, or perhaps bda practitioners simply don't care what transformations are made to the posterior (I'm not asking rhetorically - this is a possible answer)?
Oct 10	comment	Reporting contrasts between binary probability parameters in Bayesian data analysis - odds ratios or difference in probability? I realize it's pretty straightforward to do transformations on the posterior samples (odds, log odds, risk ratios, etc), the question is regarding the preferred/standard presentation.
Oct 10	asked	Reporting contrasts between binary probability parameters in Bayesian data analysis - odds ratios or difference in probability?
Sep 18	accepted	Latent variables, overparameterization and MCMC convergence in bayesian models
Sep 4	comment	Combination of variational methods and empirical Bayes not my area of the literature, but are you aware of jaakola's work on variational methods for bayesian parameter estimation? Sorry if this is old news..