Reputation
9,256
Top tag
Next privilege 10,000 Rep.
Access moderator tools
Badges
24 49
Newest
 Guru
Impact
~564k people reached

4h
comment When (if ever) is a frequentist approach substantively better than a Bayesian?
@CliffAB: Ease-of-use is important, and as you say if the results are of equal quality, why choose harder-to-use? At the same time, thinking about, making explicit, and understanding priors (not Bayesian, I mean literally the priors that every scientist, every field, and every study has) is critical to good science. Bayesian statistics is explicit and forces you to think about and understand some of these issues. To the extent that this is not merely pedantic inconvenience, it's arguably good, and so its opposite isn't slam-dunk good either.
4h
answered When (if ever) is a frequentist approach substantively better than a Bayesian?
4h
comment When (if ever) is a frequentist approach substantively better than a Bayesian?
In terms of freedom from priors: are you saying that the less you have to think about and understand your problem, the better? I know several software vendors who would like to talk to you, so you can point-n-click -- or better yet, one-click -- and have an answer to any problem you can imagine! Heck, you don't even need a problem, just feed your data into their website and they'll find all possible problems and solve them, toot sweet ! (Sorry, couldn't resist answering with a cruel straw-man-like comment.)
6h
comment How can I help ensure testing data does not leak into training data?
This question is old and has an accepted answer, but I can't get out of my head that the initial premise seems to be conflicted. If this is about learning exercises, then the best way to approach it is to let them fail and make an overfit model that subsequently doesn't predict well. (Perhaps keep a secret holdout dataset that you show them the error of their ways.) But it also sounds like you've got unqualified people making models that someone is going to take seriously and act upon and you're trying to minimize the damage at arms length. There are a LOT of subtleties, as you say.
Feb
4
awarded  Guru
Feb
3
comment How does the inclusion of an intercept change the variability of the residual?
+1. Omitted variables, misspecified models, and a whole host of issues make it hard to believe that the residual will meaningfully reflect the "manager" effect.
Feb
3
comment How does the inclusion of an intercept change the variability of the residual?
@user51972: Wow, that explanation sounds crazy. If your list of X was perfect and all-inclusive and if your model was of the correct form, this method would work I guess. Is this really widely used?
Feb
1
comment In OLS, how would log transformation of variables affect the estimation of coefficients?
I agree with Peter that this is different. At the same time, I'd ask yliu95 to think through and clarify the question. For example, "will the ... significance level of the coefficients be different": that's exactly why you would transform in the first place: the transformation would make the relationship more realistic (linear?) and therefore it should be more significant.
Jan
27
answered Customization of a standard Bell Curve
Jan
26
comment Customization of a standard Bell Curve
Are you simply wanting to customize pictures (graphs) or are you wanting to do something with them? If you want to do something with them, what do you want to do, specifically?
Jan
26
comment Are Bayes factors practically applicable?
I do not have an answer for you, but I'm pretty certain that using Bayes Factors to compare/select models is discouraged now-a-days because the comparison is very sensitive to the priors you used.
Jan
26
comment Is there “rule of 30” in machine learning
+1. How could there even be such a rule? As you point out, some kind of improvement based on a percentage of the test cases might be a reasonable rule of thumb, but an absolute number?
Jan
26
comment Can someone explain this cartoon?
You are calculating the probability of exactly one occurrence, rather than the probability of one or more occurrences. And it's more subtle than that because we're dealing with a frequentist concept here and $\alpha=0.05$ does not mean what we might want it to mean. The cartoon's point is that, in layman's terms, if you allow that your significance test can be wrong as often as 5% of the time, you shouldn't be surprised when it is wrong 5% of the time (i.e. 1 of 20). Again, frequentist concepts are tricky so technically the previous sentence may be a mis-statement, but as a concept it's fine
Jan
25
comment Can someone explain this cartoon?
Argh, I though it was a paper, but it's a slide deck. Should've looked closer. I hope you did get something out of it, since a lot of detail is left out of the slides to give the speaker room to talk. Bottom line is, even with the binomial math you did -- which isn't quite what you think it is -- the odds are almost 2:1 (0.6415 v 0.3585) that you have a problem. Would you listen to a lecture where someone started with, "the odds are 1 out of 3 that what I found might be real!" (Apply your math to your 160 regressions and the odds explode.)
Jan
25
comment Can someone explain this cartoon?
**Your professor is using a humorous cartoon to warn you about the trap you have stepped into. Don't try to argue with the cartoon instead of heading the warning. You're right to want to learn more about the problem: semanticommunity.info/@api/deki/files/30744/… **
Jan
24
comment Why use logistic regression instead of SVM?
@rolando2: Remember, this is a "Personally, I..." statement... SVMs have a variety of settings that I find over-sensitive and don't feel like I can adjust in a principled way. (And I'm not a fan of grid searches over parameter combinations.) They're not straightforward to tune, in my opinion, so perhaps "fiddly" is a better word.
Jan
24
answered Why use logistic regression instead of SVM?
Jan
23
comment Does this Monte Carlo Technique Have a Name?
+1. And as the wikipedia entry says in the first sentence, but I'll repeat here for future searches -- does the search look at comments? -- this technique is also known as a "Particle Filter". There is a: Stan State Space github project that I just found in a search, that might or might not be helpful: github.com/sinhrks/stan-statespace .
Jan
23
comment SVM Vs Neural Network Vs Random Forest classifier comparison on multi class problem
+1 All three methods are comparable. It would depend on the dataset and also on your familiarity with the method.
Jan
20
comment What is the best introductory Bayesian statistics textbook?
This is an old thread now, but I came back to +1 a new book "Statistical Rethinking. And in looking the higher-ranking answers in the thread, I think a key distinction hasn't been made: "introductory" for whom? A first course in statistics (that happens to have a Bayesian approach)? An introduction to Bayesian methods for someone with basic undergraduate (non-Bayesian) statistics classes? Or an introduction to Bayesian statistics for a practitioner of non-Bayesian statistics who has finally been persuaded that this Bayesian thing isn't a fad? Very different introductions.