d-diaz / sefs590f_winter2017 Goto Github PK

Clarify notation to distinguish Mu in likelihood and Mu0 as parameter of the distribution of Mu in prior and posterior distributions

7. Find the mean, variance, and 95% quantiles of 10000 random draws from a Poisson distribution with λ=33.32

Originally gave 95th percentile instead of top-and-bottom of 95% (0.025 to 0.975 percentile)

Should be doing log transformation of parameters, not PDF. Probability densities should be positive...

6. d. The probability that a white male will vote republican in a presidential election.
   Probability of white male voting. Value between 0 and 1, a continuous, not discrete response variable. Binomial distribution is discrete (original answer). Should be updated to Beta distribution.

Clarify distinction between standard deviation of data compared to standard deviation of mean

e. The number of individuals in four mutually exclusive income categories.
Response variable is discrete counts within mutually exclusive categories. Multinomial is more appropriate distribution.

avg height is deterministic, so can be left off of DAG. sigma cannot be negative, so change prior distribution to something other than normal.

in code box 10, 12, and 13 should be using np.prod() instead of np.sum() if using likelihood (norm.pdf), or transform to log likelihood (norm.logpdf) if you want to sum.

instead of np.std(observations), use np.std(observations, ddof=1) to correct for small sample size.

on 1.c, 2.b, 2.e, should be using distribution parameters only rather than providing means and sd to imply moment matching.