Maximum likelihood parameters deviate from posterior distributions Announcing the arrival of Valued Associate #679: Cesar Manara Planned maintenance scheduled April 23, 2019 at 23:30 UTC (7:30pm US/Eastern)lme() and lmer() giving conflicting resultsWhy is MCMC needed when estimating a parameter using MAPGiven MCMC samples, what are the options for estimating posterior of parameters?Maximizing likelihood versus MCMC sampling: Comparing Parameters and DevianceModelling parameters in maximum likelihoodMarkov chain Monte Carlo (MCMC) for Maximum Likelihood Estimation (MLE)What is the common criterion to decide the performance of prior selection in MCMCConnection between MCMC and Optimization for Inverse/Parameter-Estmation ProblemsFailure of Maximum Likelihood EstimationDo posterior probability values from an MCMC analysis have any use?Monte Carlo maximum likelihood vs Bayesian inferenceExample of maximum a posteriori that does not match the mean of a marginalized posterior
What's the difference between using dependency injection with a container and using a service locator?
Why aren't these two solutions equivalent? Combinatorics problem
Recursive calls to a function - why is the address of the parameter passed to it lowering with each call?
A journey... into the MIND
How to know or convert AREA, PERIMETER units in QGIS
Assertions In A Mock Callout Test
Why doesn't the university give past final exams' answers?
Why do people think Winterfell crypts is the safest place for women, children & old people?
Marquee sign letters
Is there a verb for listening stealthily?
Coin Game with infinite paradox
Does Prince Arnaud cause someone holding the Princess to lose?
How to create a command for the "strange m" symbol in latex?
Can this water damage be explained by lack of gutters and grading issues?
Why does BitLocker not use RSA?
How to calculate density of unknown planet?
Putting Ant-Man on house arrest
Does traveling In The United States require a passport or can I use my green card if not a US citizen?
Why does my GNOME settings mention "Moto C Plus"?
Why is one lightbulb in a string illuminated?
How is an IPA symbol that lacks a name (e.g. ɲ) called?
How to ask rejected full-time candidates to apply to teach individual courses?
Why did Israel vote against lifting the American embargo on Cuba?
What *exactly* is electrical current, voltage, and resistance?
Maximum likelihood parameters deviate from posterior distributions
Announcing the arrival of Valued Associate #679: Cesar Manara
Planned maintenance scheduled April 23, 2019 at 23:30 UTC (7:30pm US/Eastern)lme() and lmer() giving conflicting resultsWhy is MCMC needed when estimating a parameter using MAPGiven MCMC samples, what are the options for estimating posterior of parameters?Maximizing likelihood versus MCMC sampling: Comparing Parameters and DevianceModelling parameters in maximum likelihoodMarkov chain Monte Carlo (MCMC) for Maximum Likelihood Estimation (MLE)What is the common criterion to decide the performance of prior selection in MCMCConnection between MCMC and Optimization for Inverse/Parameter-Estmation ProblemsFailure of Maximum Likelihood EstimationDo posterior probability values from an MCMC analysis have any use?Monte Carlo maximum likelihood vs Bayesian inferenceExample of maximum a posteriori that does not match the mean of a marginalized posterior
.everyoneloves__top-leaderboard:empty,.everyoneloves__mid-leaderboard:empty,.everyoneloves__bot-mid-leaderboard:empty margin-bottom:0;
$begingroup$
I have a likelihood function $mathcalL(d | theta)$ for the probability of my data $d$ given some model parameters $theta in mathbfR^N$, which I would like to estimate. Assuming flat priors on the parameters, the likelihood is proportional to the posterior probability. I use an MCMC method to sample this probability.
Looking at the resulting converged chain, I find that the maximum likelihood parameters are not consistent with the posterior distributions. For example, the marginalized posterior probability distribution for one of the parameters might be $theta_0 sim N(mu=0, sigma^2=1)$, while the value of $theta_0$ at the maximum likelihood point is $theta_0^ML approx 4$, essentially being almost the maximum value of $theta_0$ traversed by the MCMC sampler.
This is an illustrative example, not my actual results. The real distributions are far more complicated, but some of the ML parameters have similarly unlikely p-values in their respective posterior distributions. Note that some of my parameters are bounded (e.g. $0 leq theta_1 leq 1$); within the bounds, the priors are always uniform.
My questions are:
Is such a deviation a problem per se? Obviously I do not expect the ML parameters to exactly coincide which the maxima of each of their marginalized posterior distributions, but intuitively it feels like they should also not be found deep in the tails. Does this deviation automatically invalidate my results?
Whether this is necessarily problematic or not, could it be symptomatic of specific pathologies at some stage of the data analysis? For example, is it possible to make any general statement about whether such a deviation could be induced by an improperly converged chain, an incorrect model, or excessively tight bounds on the parameters?
bayesian maximum-likelihood optimization inference mcmc
$endgroup$
add a comment |
$begingroup$
I have a likelihood function $mathcalL(d | theta)$ for the probability of my data $d$ given some model parameters $theta in mathbfR^N$, which I would like to estimate. Assuming flat priors on the parameters, the likelihood is proportional to the posterior probability. I use an MCMC method to sample this probability.
Looking at the resulting converged chain, I find that the maximum likelihood parameters are not consistent with the posterior distributions. For example, the marginalized posterior probability distribution for one of the parameters might be $theta_0 sim N(mu=0, sigma^2=1)$, while the value of $theta_0$ at the maximum likelihood point is $theta_0^ML approx 4$, essentially being almost the maximum value of $theta_0$ traversed by the MCMC sampler.
This is an illustrative example, not my actual results. The real distributions are far more complicated, but some of the ML parameters have similarly unlikely p-values in their respective posterior distributions. Note that some of my parameters are bounded (e.g. $0 leq theta_1 leq 1$); within the bounds, the priors are always uniform.
My questions are:
Is such a deviation a problem per se? Obviously I do not expect the ML parameters to exactly coincide which the maxima of each of their marginalized posterior distributions, but intuitively it feels like they should also not be found deep in the tails. Does this deviation automatically invalidate my results?
Whether this is necessarily problematic or not, could it be symptomatic of specific pathologies at some stage of the data analysis? For example, is it possible to make any general statement about whether such a deviation could be induced by an improperly converged chain, an incorrect model, or excessively tight bounds on the parameters?
bayesian maximum-likelihood optimization inference mcmc
$endgroup$
add a comment |
$begingroup$
I have a likelihood function $mathcalL(d | theta)$ for the probability of my data $d$ given some model parameters $theta in mathbfR^N$, which I would like to estimate. Assuming flat priors on the parameters, the likelihood is proportional to the posterior probability. I use an MCMC method to sample this probability.
Looking at the resulting converged chain, I find that the maximum likelihood parameters are not consistent with the posterior distributions. For example, the marginalized posterior probability distribution for one of the parameters might be $theta_0 sim N(mu=0, sigma^2=1)$, while the value of $theta_0$ at the maximum likelihood point is $theta_0^ML approx 4$, essentially being almost the maximum value of $theta_0$ traversed by the MCMC sampler.
This is an illustrative example, not my actual results. The real distributions are far more complicated, but some of the ML parameters have similarly unlikely p-values in their respective posterior distributions. Note that some of my parameters are bounded (e.g. $0 leq theta_1 leq 1$); within the bounds, the priors are always uniform.
My questions are:
Is such a deviation a problem per se? Obviously I do not expect the ML parameters to exactly coincide which the maxima of each of their marginalized posterior distributions, but intuitively it feels like they should also not be found deep in the tails. Does this deviation automatically invalidate my results?
Whether this is necessarily problematic or not, could it be symptomatic of specific pathologies at some stage of the data analysis? For example, is it possible to make any general statement about whether such a deviation could be induced by an improperly converged chain, an incorrect model, or excessively tight bounds on the parameters?
bayesian maximum-likelihood optimization inference mcmc
$endgroup$
I have a likelihood function $mathcalL(d | theta)$ for the probability of my data $d$ given some model parameters $theta in mathbfR^N$, which I would like to estimate. Assuming flat priors on the parameters, the likelihood is proportional to the posterior probability. I use an MCMC method to sample this probability.
Looking at the resulting converged chain, I find that the maximum likelihood parameters are not consistent with the posterior distributions. For example, the marginalized posterior probability distribution for one of the parameters might be $theta_0 sim N(mu=0, sigma^2=1)$, while the value of $theta_0$ at the maximum likelihood point is $theta_0^ML approx 4$, essentially being almost the maximum value of $theta_0$ traversed by the MCMC sampler.
This is an illustrative example, not my actual results. The real distributions are far more complicated, but some of the ML parameters have similarly unlikely p-values in their respective posterior distributions. Note that some of my parameters are bounded (e.g. $0 leq theta_1 leq 1$); within the bounds, the priors are always uniform.
My questions are:
Is such a deviation a problem per se? Obviously I do not expect the ML parameters to exactly coincide which the maxima of each of their marginalized posterior distributions, but intuitively it feels like they should also not be found deep in the tails. Does this deviation automatically invalidate my results?
Whether this is necessarily problematic or not, could it be symptomatic of specific pathologies at some stage of the data analysis? For example, is it possible to make any general statement about whether such a deviation could be induced by an improperly converged chain, an incorrect model, or excessively tight bounds on the parameters?
bayesian maximum-likelihood optimization inference mcmc
bayesian maximum-likelihood optimization inference mcmc
asked Apr 5 at 11:25
mgc70mgc70
484
484
add a comment |
add a comment |
2 Answers
2
active
oldest
votes
$begingroup$
With flat priors, the posterior is identical to the likelihood up to a constant. Thus
MLE (estimated with an optimizer) should be identical to the MAP (maximum a posteriori value = multivariate mode of the posterior, estimated with MCMC). If you don't get the same value, you have a problem with your sampler or optimiser.
For complex models, it is very common that the marginal modes are different from the MAP. This happens, for example, if correlations between parameters are nonlinear. This is perfectly fine, but marginal modes should therefore not be interpreted as the points of highest posterior density, and not be compared to the MLE.
In your specific case, however, I suspect that the posterior runs against the prior boundary. In this case, the posterior will be strongly asymmetric, and it doesn't make sense to interpret it in terms of mean, sd. There is no principle problem with this situation, but in practice it often hints towards model misspecification, or poorly chosen priors.
$endgroup$
add a comment |
$begingroup$
Some possible generic explanations for this perceived discrepancy, assuming of course there is no issue with code or likelihood definition or MCMC implementation or number of MCMC iterations or convergence of the likelihood maximiser (thanks, Jacob Socolar):
in large dimensions $N$, the posterior does not concentrate on the
maximum but something of a distance of order $sqrtN$ from the
mode, meaning that the largest values of the likelihood function
encountered by an MCMC sampler are often quite below the value of
the likelihood at its maximum. For instance, if the posterior is $theta|mathbf xsimmathcal N_N(0,I_N)$, $theta$ is at least at a distance $N-2sqrt2N$ from the mode, $0$.While the MAP and the MLE are indeed confounded under a flat prior, the
marginal densities of the different parameters of the model may have (marginal) modes
that are far away from the corresponding MLEs (i.e., MAPs).The MAP is a position
in the parameter space where the posterior density is highest but
this does not convey any indication of posterior weight or volume
for neighbourhoods of the MAP. A very thin spike carries no posterior weight. This is also the reason why MCMC exploration of a posterior may face difficulties in identifying the posterior mode.The fact that most parameters are bounded may lead to some
components of the MAP=MLE occurring at a boundary.
See, e.g., Druihlet and Marin (2007) for arguments on the un-Bayesian nature of MAP estimators. One is the dependence on these estimators on the dominating measure, another one being the lack of invariance under reparameterisation (unlike MLE's).
As an example of point 1 above, here is a short R code
N=100
T=1e4
lik=dis=rep(0,T)
mu=rmvnorm(1,mean=rep(0,N))
xobs=rmvnorm(1,mean=rep(0,N))
lik[1]=dmvnorm(xobs,mu,log=TRUE)
dis[1]=(xobs-mu)%*%t(xobs-mu)
for (t in 2:T)
prop=rmvnorm(1,mean=mu,sigma=diag(1/N,N))
proike=dmvnorm(xobs,prop,log=TRUE)
if (log(runif(1))<proike-lik[t-1])
mu=prop;lik[t]=proike
elselik[t]=lik[t-1]
dis[t]=(xobs-mu)%*%t(xobs-mu)
which mimics a random-walk Metropolis-Hastings sequence in dimension N=100. The value of the log-likelihood at the MAP is -91.89, but the visited likelihoods never come close:
> range(lik)
[1] -183.9515 -126.6924
which is explained by the fact that the sequence never comes near the observation:
> range(dis)
[1] 69.59714 184.11525
$endgroup$
2
$begingroup$
I'd just add that in addition to worrying about the code or likelihood definition or MCMC implementation, the OP might also worry about whether the software used to obtain the ML estimate got trapped in a local optimum. stats.stackexchange.com/questions/384528/…
$endgroup$
– Jacob Socolar
Apr 5 at 15:30
add a comment |
Your Answer
StackExchange.ready(function()
var channelOptions =
tags: "".split(" "),
id: "65"
;
initTagRenderer("".split(" "), "".split(" "), channelOptions);
StackExchange.using("externalEditor", function()
// Have to fire editor after snippets, if snippets enabled
if (StackExchange.settings.snippets.snippetsEnabled)
StackExchange.using("snippets", function()
createEditor();
);
else
createEditor();
);
function createEditor()
StackExchange.prepareEditor(
heartbeatType: 'answer',
autoActivateHeartbeat: false,
convertImagesToLinks: false,
noModals: true,
showLowRepImageUploadWarning: true,
reputationToPostImages: null,
bindNavPrevention: true,
postfix: "",
imageUploader:
brandingHtml: "Powered by u003ca class="icon-imgur-white" href="https://imgur.com/"u003eu003c/au003e",
contentPolicyHtml: "User contributions licensed under u003ca href="https://creativecommons.org/licenses/by-sa/3.0/"u003ecc by-sa 3.0 with attribution requiredu003c/au003e u003ca href="https://stackoverflow.com/legal/content-policy"u003e(content policy)u003c/au003e",
allowUrls: true
,
onDemand: true,
discardSelector: ".discard-answer"
,immediatelyShowMarkdownHelp:true
);
);
Sign up or log in
StackExchange.ready(function ()
StackExchange.helpers.onClickDraftSave('#login-link');
);
Sign up using Google
Sign up using Facebook
Sign up using Email and Password
Post as a guest
Required, but never shown
StackExchange.ready(
function ()
StackExchange.openid.initPostLogin('.new-post-login', 'https%3a%2f%2fstats.stackexchange.com%2fquestions%2f401349%2fmaximum-likelihood-parameters-deviate-from-posterior-distributions%23new-answer', 'question_page');
);
Post as a guest
Required, but never shown
2 Answers
2
active
oldest
votes
2 Answers
2
active
oldest
votes
active
oldest
votes
active
oldest
votes
$begingroup$
With flat priors, the posterior is identical to the likelihood up to a constant. Thus
MLE (estimated with an optimizer) should be identical to the MAP (maximum a posteriori value = multivariate mode of the posterior, estimated with MCMC). If you don't get the same value, you have a problem with your sampler or optimiser.
For complex models, it is very common that the marginal modes are different from the MAP. This happens, for example, if correlations between parameters are nonlinear. This is perfectly fine, but marginal modes should therefore not be interpreted as the points of highest posterior density, and not be compared to the MLE.
In your specific case, however, I suspect that the posterior runs against the prior boundary. In this case, the posterior will be strongly asymmetric, and it doesn't make sense to interpret it in terms of mean, sd. There is no principle problem with this situation, but in practice it often hints towards model misspecification, or poorly chosen priors.
$endgroup$
add a comment |
$begingroup$
With flat priors, the posterior is identical to the likelihood up to a constant. Thus
MLE (estimated with an optimizer) should be identical to the MAP (maximum a posteriori value = multivariate mode of the posterior, estimated with MCMC). If you don't get the same value, you have a problem with your sampler or optimiser.
For complex models, it is very common that the marginal modes are different from the MAP. This happens, for example, if correlations between parameters are nonlinear. This is perfectly fine, but marginal modes should therefore not be interpreted as the points of highest posterior density, and not be compared to the MLE.
In your specific case, however, I suspect that the posterior runs against the prior boundary. In this case, the posterior will be strongly asymmetric, and it doesn't make sense to interpret it in terms of mean, sd. There is no principle problem with this situation, but in practice it often hints towards model misspecification, or poorly chosen priors.
$endgroup$
add a comment |
$begingroup$
With flat priors, the posterior is identical to the likelihood up to a constant. Thus
MLE (estimated with an optimizer) should be identical to the MAP (maximum a posteriori value = multivariate mode of the posterior, estimated with MCMC). If you don't get the same value, you have a problem with your sampler or optimiser.
For complex models, it is very common that the marginal modes are different from the MAP. This happens, for example, if correlations between parameters are nonlinear. This is perfectly fine, but marginal modes should therefore not be interpreted as the points of highest posterior density, and not be compared to the MLE.
In your specific case, however, I suspect that the posterior runs against the prior boundary. In this case, the posterior will be strongly asymmetric, and it doesn't make sense to interpret it in terms of mean, sd. There is no principle problem with this situation, but in practice it often hints towards model misspecification, or poorly chosen priors.
$endgroup$
With flat priors, the posterior is identical to the likelihood up to a constant. Thus
MLE (estimated with an optimizer) should be identical to the MAP (maximum a posteriori value = multivariate mode of the posterior, estimated with MCMC). If you don't get the same value, you have a problem with your sampler or optimiser.
For complex models, it is very common that the marginal modes are different from the MAP. This happens, for example, if correlations between parameters are nonlinear. This is perfectly fine, but marginal modes should therefore not be interpreted as the points of highest posterior density, and not be compared to the MLE.
In your specific case, however, I suspect that the posterior runs against the prior boundary. In this case, the posterior will be strongly asymmetric, and it doesn't make sense to interpret it in terms of mean, sd. There is no principle problem with this situation, but in practice it often hints towards model misspecification, or poorly chosen priors.
edited Apr 8 at 11:42
answered Apr 5 at 11:54
Florian HartigFlorian Hartig
4,2591423
4,2591423
add a comment |
add a comment |
$begingroup$
Some possible generic explanations for this perceived discrepancy, assuming of course there is no issue with code or likelihood definition or MCMC implementation or number of MCMC iterations or convergence of the likelihood maximiser (thanks, Jacob Socolar):
in large dimensions $N$, the posterior does not concentrate on the
maximum but something of a distance of order $sqrtN$ from the
mode, meaning that the largest values of the likelihood function
encountered by an MCMC sampler are often quite below the value of
the likelihood at its maximum. For instance, if the posterior is $theta|mathbf xsimmathcal N_N(0,I_N)$, $theta$ is at least at a distance $N-2sqrt2N$ from the mode, $0$.While the MAP and the MLE are indeed confounded under a flat prior, the
marginal densities of the different parameters of the model may have (marginal) modes
that are far away from the corresponding MLEs (i.e., MAPs).The MAP is a position
in the parameter space where the posterior density is highest but
this does not convey any indication of posterior weight or volume
for neighbourhoods of the MAP. A very thin spike carries no posterior weight. This is also the reason why MCMC exploration of a posterior may face difficulties in identifying the posterior mode.The fact that most parameters are bounded may lead to some
components of the MAP=MLE occurring at a boundary.
See, e.g., Druihlet and Marin (2007) for arguments on the un-Bayesian nature of MAP estimators. One is the dependence on these estimators on the dominating measure, another one being the lack of invariance under reparameterisation (unlike MLE's).
As an example of point 1 above, here is a short R code
N=100
T=1e4
lik=dis=rep(0,T)
mu=rmvnorm(1,mean=rep(0,N))
xobs=rmvnorm(1,mean=rep(0,N))
lik[1]=dmvnorm(xobs,mu,log=TRUE)
dis[1]=(xobs-mu)%*%t(xobs-mu)
for (t in 2:T)
prop=rmvnorm(1,mean=mu,sigma=diag(1/N,N))
proike=dmvnorm(xobs,prop,log=TRUE)
if (log(runif(1))<proike-lik[t-1])
mu=prop;lik[t]=proike
elselik[t]=lik[t-1]
dis[t]=(xobs-mu)%*%t(xobs-mu)
which mimics a random-walk Metropolis-Hastings sequence in dimension N=100. The value of the log-likelihood at the MAP is -91.89, but the visited likelihoods never come close:
> range(lik)
[1] -183.9515 -126.6924
which is explained by the fact that the sequence never comes near the observation:
> range(dis)
[1] 69.59714 184.11525
$endgroup$
2
$begingroup$
I'd just add that in addition to worrying about the code or likelihood definition or MCMC implementation, the OP might also worry about whether the software used to obtain the ML estimate got trapped in a local optimum. stats.stackexchange.com/questions/384528/…
$endgroup$
– Jacob Socolar
Apr 5 at 15:30
add a comment |
$begingroup$
Some possible generic explanations for this perceived discrepancy, assuming of course there is no issue with code or likelihood definition or MCMC implementation or number of MCMC iterations or convergence of the likelihood maximiser (thanks, Jacob Socolar):
in large dimensions $N$, the posterior does not concentrate on the
maximum but something of a distance of order $sqrtN$ from the
mode, meaning that the largest values of the likelihood function
encountered by an MCMC sampler are often quite below the value of
the likelihood at its maximum. For instance, if the posterior is $theta|mathbf xsimmathcal N_N(0,I_N)$, $theta$ is at least at a distance $N-2sqrt2N$ from the mode, $0$.While the MAP and the MLE are indeed confounded under a flat prior, the
marginal densities of the different parameters of the model may have (marginal) modes
that are far away from the corresponding MLEs (i.e., MAPs).The MAP is a position
in the parameter space where the posterior density is highest but
this does not convey any indication of posterior weight or volume
for neighbourhoods of the MAP. A very thin spike carries no posterior weight. This is also the reason why MCMC exploration of a posterior may face difficulties in identifying the posterior mode.The fact that most parameters are bounded may lead to some
components of the MAP=MLE occurring at a boundary.
See, e.g., Druihlet and Marin (2007) for arguments on the un-Bayesian nature of MAP estimators. One is the dependence on these estimators on the dominating measure, another one being the lack of invariance under reparameterisation (unlike MLE's).
As an example of point 1 above, here is a short R code
N=100
T=1e4
lik=dis=rep(0,T)
mu=rmvnorm(1,mean=rep(0,N))
xobs=rmvnorm(1,mean=rep(0,N))
lik[1]=dmvnorm(xobs,mu,log=TRUE)
dis[1]=(xobs-mu)%*%t(xobs-mu)
for (t in 2:T)
prop=rmvnorm(1,mean=mu,sigma=diag(1/N,N))
proike=dmvnorm(xobs,prop,log=TRUE)
if (log(runif(1))<proike-lik[t-1])
mu=prop;lik[t]=proike
elselik[t]=lik[t-1]
dis[t]=(xobs-mu)%*%t(xobs-mu)
which mimics a random-walk Metropolis-Hastings sequence in dimension N=100. The value of the log-likelihood at the MAP is -91.89, but the visited likelihoods never come close:
> range(lik)
[1] -183.9515 -126.6924
which is explained by the fact that the sequence never comes near the observation:
> range(dis)
[1] 69.59714 184.11525
$endgroup$
2
$begingroup$
I'd just add that in addition to worrying about the code or likelihood definition or MCMC implementation, the OP might also worry about whether the software used to obtain the ML estimate got trapped in a local optimum. stats.stackexchange.com/questions/384528/…
$endgroup$
– Jacob Socolar
Apr 5 at 15:30
add a comment |
$begingroup$
Some possible generic explanations for this perceived discrepancy, assuming of course there is no issue with code or likelihood definition or MCMC implementation or number of MCMC iterations or convergence of the likelihood maximiser (thanks, Jacob Socolar):
in large dimensions $N$, the posterior does not concentrate on the
maximum but something of a distance of order $sqrtN$ from the
mode, meaning that the largest values of the likelihood function
encountered by an MCMC sampler are often quite below the value of
the likelihood at its maximum. For instance, if the posterior is $theta|mathbf xsimmathcal N_N(0,I_N)$, $theta$ is at least at a distance $N-2sqrt2N$ from the mode, $0$.While the MAP and the MLE are indeed confounded under a flat prior, the
marginal densities of the different parameters of the model may have (marginal) modes
that are far away from the corresponding MLEs (i.e., MAPs).The MAP is a position
in the parameter space where the posterior density is highest but
this does not convey any indication of posterior weight or volume
for neighbourhoods of the MAP. A very thin spike carries no posterior weight. This is also the reason why MCMC exploration of a posterior may face difficulties in identifying the posterior mode.The fact that most parameters are bounded may lead to some
components of the MAP=MLE occurring at a boundary.
See, e.g., Druihlet and Marin (2007) for arguments on the un-Bayesian nature of MAP estimators. One is the dependence on these estimators on the dominating measure, another one being the lack of invariance under reparameterisation (unlike MLE's).
As an example of point 1 above, here is a short R code
N=100
T=1e4
lik=dis=rep(0,T)
mu=rmvnorm(1,mean=rep(0,N))
xobs=rmvnorm(1,mean=rep(0,N))
lik[1]=dmvnorm(xobs,mu,log=TRUE)
dis[1]=(xobs-mu)%*%t(xobs-mu)
for (t in 2:T)
prop=rmvnorm(1,mean=mu,sigma=diag(1/N,N))
proike=dmvnorm(xobs,prop,log=TRUE)
if (log(runif(1))<proike-lik[t-1])
mu=prop;lik[t]=proike
elselik[t]=lik[t-1]
dis[t]=(xobs-mu)%*%t(xobs-mu)
which mimics a random-walk Metropolis-Hastings sequence in dimension N=100. The value of the log-likelihood at the MAP is -91.89, but the visited likelihoods never come close:
> range(lik)
[1] -183.9515 -126.6924
which is explained by the fact that the sequence never comes near the observation:
> range(dis)
[1] 69.59714 184.11525
$endgroup$
Some possible generic explanations for this perceived discrepancy, assuming of course there is no issue with code or likelihood definition or MCMC implementation or number of MCMC iterations or convergence of the likelihood maximiser (thanks, Jacob Socolar):
in large dimensions $N$, the posterior does not concentrate on the
maximum but something of a distance of order $sqrtN$ from the
mode, meaning that the largest values of the likelihood function
encountered by an MCMC sampler are often quite below the value of
the likelihood at its maximum. For instance, if the posterior is $theta|mathbf xsimmathcal N_N(0,I_N)$, $theta$ is at least at a distance $N-2sqrt2N$ from the mode, $0$.While the MAP and the MLE are indeed confounded under a flat prior, the
marginal densities of the different parameters of the model may have (marginal) modes
that are far away from the corresponding MLEs (i.e., MAPs).The MAP is a position
in the parameter space where the posterior density is highest but
this does not convey any indication of posterior weight or volume
for neighbourhoods of the MAP. A very thin spike carries no posterior weight. This is also the reason why MCMC exploration of a posterior may face difficulties in identifying the posterior mode.The fact that most parameters are bounded may lead to some
components of the MAP=MLE occurring at a boundary.
See, e.g., Druihlet and Marin (2007) for arguments on the un-Bayesian nature of MAP estimators. One is the dependence on these estimators on the dominating measure, another one being the lack of invariance under reparameterisation (unlike MLE's).
As an example of point 1 above, here is a short R code
N=100
T=1e4
lik=dis=rep(0,T)
mu=rmvnorm(1,mean=rep(0,N))
xobs=rmvnorm(1,mean=rep(0,N))
lik[1]=dmvnorm(xobs,mu,log=TRUE)
dis[1]=(xobs-mu)%*%t(xobs-mu)
for (t in 2:T)
prop=rmvnorm(1,mean=mu,sigma=diag(1/N,N))
proike=dmvnorm(xobs,prop,log=TRUE)
if (log(runif(1))<proike-lik[t-1])
mu=prop;lik[t]=proike
elselik[t]=lik[t-1]
dis[t]=(xobs-mu)%*%t(xobs-mu)
which mimics a random-walk Metropolis-Hastings sequence in dimension N=100. The value of the log-likelihood at the MAP is -91.89, but the visited likelihoods never come close:
> range(lik)
[1] -183.9515 -126.6924
which is explained by the fact that the sequence never comes near the observation:
> range(dis)
[1] 69.59714 184.11525
edited Apr 8 at 11:40
answered Apr 5 at 11:57
Xi'anXi'an
59.8k897368
59.8k897368
2
$begingroup$
I'd just add that in addition to worrying about the code or likelihood definition or MCMC implementation, the OP might also worry about whether the software used to obtain the ML estimate got trapped in a local optimum. stats.stackexchange.com/questions/384528/…
$endgroup$
– Jacob Socolar
Apr 5 at 15:30
add a comment |
2
$begingroup$
I'd just add that in addition to worrying about the code or likelihood definition or MCMC implementation, the OP might also worry about whether the software used to obtain the ML estimate got trapped in a local optimum. stats.stackexchange.com/questions/384528/…
$endgroup$
– Jacob Socolar
Apr 5 at 15:30
2
2
$begingroup$
I'd just add that in addition to worrying about the code or likelihood definition or MCMC implementation, the OP might also worry about whether the software used to obtain the ML estimate got trapped in a local optimum. stats.stackexchange.com/questions/384528/…
$endgroup$
– Jacob Socolar
Apr 5 at 15:30
$begingroup$
I'd just add that in addition to worrying about the code or likelihood definition or MCMC implementation, the OP might also worry about whether the software used to obtain the ML estimate got trapped in a local optimum. stats.stackexchange.com/questions/384528/…
$endgroup$
– Jacob Socolar
Apr 5 at 15:30
add a comment |
Thanks for contributing an answer to Cross Validated!
- Please be sure to answer the question. Provide details and share your research!
But avoid …
- Asking for help, clarification, or responding to other answers.
- Making statements based on opinion; back them up with references or personal experience.
Use MathJax to format equations. MathJax reference.
To learn more, see our tips on writing great answers.
Sign up or log in
StackExchange.ready(function ()
StackExchange.helpers.onClickDraftSave('#login-link');
);
Sign up using Google
Sign up using Facebook
Sign up using Email and Password
Post as a guest
Required, but never shown
StackExchange.ready(
function ()
StackExchange.openid.initPostLogin('.new-post-login', 'https%3a%2f%2fstats.stackexchange.com%2fquestions%2f401349%2fmaximum-likelihood-parameters-deviate-from-posterior-distributions%23new-answer', 'question_page');
);
Post as a guest
Required, but never shown
Sign up or log in
StackExchange.ready(function ()
StackExchange.helpers.onClickDraftSave('#login-link');
);
Sign up using Google
Sign up using Facebook
Sign up using Email and Password
Post as a guest
Required, but never shown
Sign up or log in
StackExchange.ready(function ()
StackExchange.helpers.onClickDraftSave('#login-link');
);
Sign up using Google
Sign up using Facebook
Sign up using Email and Password
Post as a guest
Required, but never shown
Sign up or log in
StackExchange.ready(function ()
StackExchange.helpers.onClickDraftSave('#login-link');
);
Sign up using Google
Sign up using Facebook
Sign up using Email and Password
Sign up using Google
Sign up using Facebook
Sign up using Email and Password
Post as a guest
Required, but never shown
Required, but never shown
Required, but never shown
Required, but never shown
Required, but never shown
Required, but never shown
Required, but never shown
Required, but never shown
Required, but never shown