# Bayesian as modeling vs Bayesian as computation

Sep 23, 2019       2 mins     Tag: casual, modeling
I came up with this random thought when I was sitting in Andrew’s class this afternoon (as a TA). In order to emphasize why Bayesian is different, Andrew listed a few alternatives for data analysis,...

# Napoleon's March is to a Dirichlet process what a boxplot to a Gaussian distribution

Sep 21, 2019       2 mins     Tag: visualization
I saw a visualization of wealth distribution in the 2019 Credit Suisse Global Wealth Report (https://www.credit-suisse.com/about-us/en/reports-research/global-wealth-report.html)

# Division of random variables

Sep 18, 2019       1 min Tag: casual
This is trivial math question but it once bothers me in the construction of Dirichlet process using stick breaking process. The context is that if $V \sim beta(1, M)$ and $\theta \sim Dirichlet (\bar \alpha)$...

# How to elegantly embed pdf image into markdown

Aug 05, 2019       2 mins     Tag: zombie
No you cannot

# Sharpe ratios

Aug 01, 2019       1 mins     Tag: zombie
This is a jitt question in Andrew’s class: to (by hand) approximate the standard deviation of X/Y if X is N(5,1) and Y is N(1,3).

# The Curse and the Blessing But Mostly the Blessing of Multimodality

Jul 30, 2019       18 mins     Tag: computation
What if I run 4 chains and find $\hat R >>1$

# ABC, Model Misspecification, and Bimodality

Jun 18, 2019       5 mins     Tag: computation
Like importance sampling, ABC is in principle immune to metastability that MCMC has to suffer, but ABC is also problematic as we have no idea what it will converge to when the model is misspecified.

# non-parametric MLE

Jun 17, 2019       1 mins     Tag: zombies
When I talked to someone about the old proof the invariance of odds ratio in a respropective sampling, I mentioned the estimation of $q(x)$ is achieved by its non-parametric MLE or its empirical distribution (see...

# Immunity to Domain Adaption, and Invariance Under Reweighting

Jun 14, 2019       9 mins     Tag: modeling decision
Some loss functions are invariant under domain adaption, which suggests we can indeed learn the optimal model of the population from a non-representative sample without sacrifice from the extrapolation.

# Should I reweight a case-control study?

Jun 01, 2019       14 mins     Tag: modeling causal
The odds ratio from a case-control study is exactly the same as in a cohort study, therefore I could fit a retrospective logistic regression as if it is prospective and report its MLE or Bayesian posterior distribution. But considering the sampling distribution shift, should I reweight it regardless?

# Automated Predictive Evaluation of Causal Inference Competitions

May 30, 2019       4 mins     Tag: causal
We can evaluate causal inference methods through simulations and based on predictive performance of ATE, but is it enough?

# The Optimal Transport and Importance Sampling

May 22, 2019       4 mins     Tag: ad-hoc computation
Most statisticians-orientated introduction to optimal transport theory starts with the motivation: we can always obtain a one-D distribution $f(\theta)$ through its inverse cdf transformation

# Affordable Bayesian Neural Networks

May 20, 2019       10 mins     Tag: computation
We wrote a new paper on approximate Bayesian inference in deep network, with a similiar computation cost as point estimation.