Reputation
113,049
Next tag badge:
273/100 score
19/20 answers
Badges
10 201 407
Newest
 Enlightened
Impact
~3.2m people reached

1m
comment Anderson-Darling test for normality with estimated parameters
Re the edits (which are very helpful and interesting): the bottom line is that a test of normality will still accomplish little. It appears that what you need to do is characterize the distributions you have, rather than merely to demonstrate they are not normal.
1h
awarded  Enlightened
1h
awarded  Nice Answer
14h
comment Is time in linear regression a categorical or continuous variable?
Time of day is not "ordinal" in any standard sense, because 0 comes right after 23! It clearly is a truncated circular variable; this observation alone indicates there are more parsimonious solutions to consider before one uses 24 dummy variables.
15h
awarded  Revival
16h
reviewed Close Algorithm to a Mathematical model Conversion (need mathematical expression for the below)
16h
comment Distinguishing two Bernoulli random sources
I think you will find some good things by using the terms I gave in my first comment for a search, such as likelihood ratio test simple hypothesis. The key idea is encapsulated in the Neyman-Pearson lemma, q.v.. I wrote down these weights immediately by supposing the sample distributions of the sums would be approximately Normal: $np$ and $nq$ will be their means and $np(1-p)$ and $nq(1-q)$ their variances.
16h
answered Derivation of the equation of ridge regression
17h
comment Why don't log-likelihoods lead to log(0)?
Although you haven't disclosed any details of your calculations, it certainly is not the case that $P(X=1|R)$ is determined by means of any kind of estimate: it is given by your model and as such will be a function of one or more parameters whose values you do not know (for otherwise you wouldn't make them parameters in the first place). But here we are discussing what a likelihood is, when there already are fine discussions elsewhere on this site. I refer you to them for more information.
17h
reviewed Close How to analyse the accuracy and standard deviation of a neural network in matlab toolbox?
17h
reviewed Close Help with Data Analysis - using Minitab
17h
comment Help with Data Analysis - using Minitab
If the responses are indeed computed from the inputs, there does not seem to be a statistical problem here: you have complete and accurate knowledge of how the responses are related to the inputs. What, then, would "most significant" mean in this context?
17h
reviewed Leave Open How to separate “rates” with TukeyHSD?
17h
revised How to separate “rates” with TukeyHSD?
added 72 characters in body
17h
comment Which one is the correct formula of confidence interval of variance?
+1 An illustrated answer is usually worth reading. Welcome to our site!
17h
revised Which one is the correct formula of confidence interval of variance?
added 16 characters in body
17h
comment Why don't log-likelihoods lead to log(0)?
At a glance, that equation does not look like it's intended to be used in implementing anything. The authors immediately reduce it to the much simpler equation (11).
17h
comment Anderson-Darling test for normality with estimated parameters
The A-D, correctly applied, works just fine when the parameters are estimated: see the Wikipedia article at en.wikipedia.org/wiki/…. It's difficult to see why this test would be terribly useful for quality control. What kinds of manufacturing processes have perfectly normal deviations?
17h
comment Distinguishing two Bernoulli random sources
Why should it be? We're talking about relative weights in a weighted average of $p$ and $q$, not the actual threshold $\gamma$. For instance, when $q=1$ the relative weights are $0$ and $\sqrt{p(1-p)}$, which would make $\gamma=1$.
17h
comment Why don't log-likelihoods lead to log(0)?
You don't. You appear to have confounded the likelihood with your own data-based estimates of the probabilities. Perhaps you might benefit from reviewing some concepts of likelihood, such as the thread at stats.stackexchange.com/questions/2641.