Henry

8,255 reputation

1720

bio	website
	location
	age	50
visits	member for	2 years, 3 months
	seen	5 hours ago
stats	profile views	478

	8,255 reputation	bio	website		visits	member for	2 years, 3 months
1720 badges		location			seen	5 hours ago

summary answers questions tags badges favorites bounties reputation activity

665 Actions

suggestions reviews revisions comments badges posts accepts all

5h	comment	Statistics Jokes That link has since moved to se16.info/hgb/statjoke.htm
Apr 26	comment	1sigma error on the mean @user603: I would not regard a lower bound for the standard deviation which is usually very weak as a connection
Apr 26	comment	Reshaping a distribution $Y = m^ \prime - {\frac {\sigma^\prime} \sigma}(X - m)$, i.e. $Y = 2m^ \prime - Z$, has the same mean and standard deviation
Apr 26	awarded	Nice Answer
Apr 8	comment	The probability of getting one variation of consecutive outcomes in Bernoulli trials before another @George: whuber's comment makes the point, and gives a better method of getting the same result. You need a transition matrix which shows that if you are at HHH then the probability of going to HHH or HHT is each 0.5. If you are at HHT then the probability of going to HTT or HTHH is each 0.5. If you are at HTHH then you stay there with probability 1. And so on. Wikipedia has articles on Markov processes and transition matrices
Apr 8	answered	The probability of getting one variation of consecutive outcomes in Bernoulli trials before another
Mar 22	comment	How I can deal with too many variables in training a data set? How many survey responses do you have?
Jan 28	awarded	Yearling
Jan 17	comment	KL divergence calculation @Legend: the probability of choosing $0.05$ from $t1$ was $2/5$ and of choosing $0.05$ from $t2$ was $0$, and of choosing $0.05$ from $M$ was $2/10 = \frac12(2/5+0)$
Jan 16	comment	KL divergence calculation @Legend The original question was for $t1 = 0.4, 0.2, 0.3, 0.05, 0.05$ and $t2 = 0.3, 0, 0.4, 0.1, 0.2$ so $M$ was simply the two them combined as a multiset.
Jan 15	comment	Logit versus Probit Though they produce very different parameters, the fitted curves are usually so close that they often cannot be meaningfully distinguished unless there is something unusual happening in the tails.
Jan 12	comment	Mathematical basis for conditional probability A small clue is that the odds $P[A] / P[\bar{A}]$ can exceed $1$ while the conditional probability $P[A\|B] = P[A,B]/P[B]$ cannot
Jan 12	answered	Residual plot with a slope of 1
Jan 11	comment	Wigan scores after 30 minutes. Calculate home, away and draw in percentage terms with poisson regression @OP: If Wigan have scored a goal after 30 minutes then, for Wigan to win overall, the position in the remaining 60 minutes of the match must be "Wigan score at least as many goals as City" which when added to Wigan's existing lead means "Wigan scores more goals than City".
Jan 11	answered	Wigan scores after 30 minutes. Calculate home, away and draw in percentage terms with poisson regression
Jan 10	comment	confidence interval for 2 sample t test You may also want to use assumptions about normal distributions and variances to construct your confidence interval
Jan 3	revised	Conditional expectation of discrete variables added 86 characters in body
Jan 3	answered	Conditional expectation of discrete variables
Jan 2	comment	Central Moments of Symmetric Distributions What is the difference between $a$ and $u$?
Dec 28	comment	Optimizing matching of players in a tournament round If you have a rating for each individual, then you could match the top pair, the next pair, etc. This will not be far from optimal.