Reputation
800
Top tag
Next privilege 1,000 Rep.
Create tags
Badges
5 18
Impact
~22k people reached

Mar
26
comment MCMC/E-M limitations?MCMC over E-M?
MCMC does not necessarily faster than EM. EM gives you a point estimate while MCMC would give you the full posterior. For your question: As long as you derive a valid MCMC algorithm for the models you mentioned (for example if you can derive a Gibbs sampler, or a valid MH sampler), then you don't have to prove any convergence. The notion of "convergence" is not well-defined in MCMC by the way...
Feb
8
awarded  Popular Question
Jan
24
comment Which is the difference between $P(A \mid B)$ and $P(A=t \mid B)$ in a Bayesian Network?
If $A$ and $B$ are both random variables, to reason about $P(A=t | B)$, you have to fix $B = b$. For example $P(A | B=b)$ could represent a probability mass function, and $P(A=t|B=b)$ is probability that $A$ being $t$ conditioned on a specific value of $B$.
Jan
18
awarded  Teacher
Jan
16
answered What does it mean u(dx) in the Fourier transform of a probability measure u?
Jan
2
comment Fourier Transform of Constant Function
Dirichlet conditions (including integrability) are sufficient, not necessary.
Dec
31
awarded  Popular Question
Dec
12
awarded  Caucus
Nov
27
awarded  Popular Question
Nov
17
comment Finite state Markov chain
A finite state (discrete-time) Markov chain is a sequence of random variables $X_1,X_2,\ldots$ all of which take values from a finite set $\mathcal{X}$.
Nov
7
awarded  Yearling
Oct
23
awarded  Popular Question
Oct
15
comment Maximum likelihood estimate of $N$ (trials) in Binomial
For fixed $x$ ($x=4$), we can write $p(x=4 | \pi,N) = \mathcal{L}(N,\pi)$. So $\pi^*$ is a maximiser for a fixed $N$, I still can not see that why this is true. Shouldn't we show that $N$ is unique in some way? May be if I plot $\mathcal{L}$ wrt $(N,\pi)$, that would shed some light.
Oct
15
revised Maximum likelihood estimate of $N$ (trials) in Binomial
edited body
Oct
14
asked Maximum likelihood estimate of $N$ (trials) in Binomial
Sep
17
awarded  Popular Question
Jul
2
awarded  Curious
Jul
2
awarded  Inquisitive
Jun
10
awarded  Notable Question
Jun
9
comment Convergence in probability of iid normal random variables
OK since I respect you a lot, I want to check :) Thanks a lot. I calculated again and I think that it is probably correct.