Mike Dunlavey

453 reputation

210

bio	website	en.wikipedia.org/wiki/…
	location	United States
	age	68
visits	member for	2 years, 8 months
	seen	2 days ago
stats	profile views	54

BS Mechanical Engr.
PhD CS(AI)
CS Prof (4yr)
Numerous consulting jobs.
15 yr at http://www.pharsight.com
Published book on CS & several articles
4 kids, 2 grand
Pilot(student)

	453 reputation	bio	website	en.wikipedia.org/wiki/…	visits	member for	2 years, 8 months
210 badges		location	United States		seen	2 days ago

summary answers questions tags badges favorites bounties reputation activity

38 Comments

suggestions reviews revisions comments badges posts accepts all

Apr 27	comment	If I randomly execute tasks working on some subset what can I say about total coverage of the set over some time period? I'm having trouble understanding the question. What exactly is the process?
Apr 23	comment	Hamming distance for strings with different length Here's a search for Levenshtein on stackoverflow.
Mar 17	comment	Was Hitler a “black swan” event? + Yeah, the conditions included a humiliated nation undergoing hyperinflation, with a highly polarized government with indecisive leaders, incapable of deciding anything, essentially creating a big power vacuum.
Feb 28	comment	Full information maximum likelihood for missing data in R Have you considered WinBugs? It handles missing data in a beautifully natural way.
Feb 7	comment	Providing variance measures for speedup ratios I would do it on a log scale.
Feb 4	comment	Standard error of estimates of covariance parameterized in tems of cholesky I'm not sure, but can you just take the diagonal elements of the Cholesky decomposition as your standard errors?
Jan 16	comment	What is intuition behind beta distribution? + NP. You put it nicely.
Jan 16	comment	What is intuition behind beta distribution? + I like your explanation of how you update the distribution when you have more data.
Nov 28	comment	How do you create a multivariate distribution with both continuous and discrete data? Can't you just break it into stages: first sample the categorical variables using a decision tree, then at each leaf of the tree sample your multivariate normal (or T, whatever). That's what we do in sampling demographic covariates for clinical trial simulation.
Nov 27	comment	Is such transformed beta a known distribution? Your answer allowed me to put the end on this answer.
Nov 21	comment	How representative is Poisson distribution of the distribution of events in reality? + for the hanging rootogram!
Nov 19	comment	Is such transformed beta a known distribution? Now I get the BetaPrime. Thx.
Nov 19	comment	Is such transformed beta a known distribution? + I should have seen the symmetry. I still don't follow your BetaPrime argument (certainly due to my rusty algebra), but here is discussed the inverse of a Beta, which tells me most of what I need. Thanks.
May 30	comment	Variation on binomial experiment @StéphaneLaurent: I was trying to assume all colors were in the urn in equal proportion, but I will look into what you recommend. Thank you.
May 29	comment	Variation on binomial experiment Right, with 5 or 10 I'm on solid ground. Even with S=2 out of N, it seems to give a good distribution for F. Anyway, thanks for giving it a shot.
May 29	comment	Variation on binomial experiment Yeah, this is a usage where I had hoped to fully understand the distribution of $F$ even as $N$ gets small. As it is I'm using $\beta$, and it works nicely, but I have this nagging feeling I'm making a mistake.
May 29	comment	Variation on binomial experiment + Thanks, but here's what bothers me about that. Suppose $S=N=1$. Then $F\sim\beta(2,1)$ whose mode is 1 and mean is 2/3. That's counterintuitive because it could be that a large fraction of the stones might all have different colors, so I really just picked one of those colors at random.
May 29	comment	Variation on binomial experiment @whuber: One at a time and replaced each time.
Jan 25	comment	Relationship between Binomial and Beta distributions It's a little late, but I finally got time to sit down and re-create your argument. The key was "multinomial coefficient". I had tried figuring it out using plain old binomial coefficients and I was getting all balled up. Thanks again for a nice answer.
Nov 21	comment	Relationship between Binomial and Beta distributions @whuber: Thanks. You're right. Something more useful is the expected value. As far as priors go, I point out that if you only see something once, it doesn't tell you much unless you happen to know the program is in an infinite (or exceedingly long) loop.