11,143 reputation
31945
bio website
location England
age
visits member for 1 year, 11 months
seen 7 mins ago

22m
comment Split Factor Levels Or Not In Variable Selection
You could combine levels based on prior knowledge. Or try group lasso, (Yuan & Lin (2006), JRSS B, 68, 1), which, while it doesn't collapse levels, still applies shrinkage (& can exclude a whole category at once). Or if you're set on collapsing levels, Gertheiss & Tutz (2010), Ann. Appl. Stat.,4,4 suggest a penalty based on all pairwise differences between levels.
19h
reviewed Close PCA- does KMO need to be over 0.5?
20h
comment What is a hazard rate?
@user7340: (1) Imagine when you die you're immediately resurrected, without being rejuvenated, & are once again at risk. It's not necessary to believe that to be a realistic scenario to employ the concept of hazard. (2) The probability at birth that you die between any specified ages is the difference in the values of the lifetime distribution function at those ages; the probability that you die in the next year is the difference in the values of the lifetime distribution functions now & one year from now divided by the value of the survival function now; the probability that you die is 1.
21h
comment Why can't the chi squared test be calculated with a row of zeros?
The question should be "Why is the on-line chi-square calculator calculating the chi-square statistic from this data?". Try it by hand & you'll see the problem.
22h
revised Interaction between a dummy variable and a variable with a quadratic form
edited title
22h
revised Interaction between a dummy variable and a variable with a quadratic form
deleted 4 characters in body
22h
revised Interaction between a dummy variable and a variable with a quadratic form
deleted 22 characters in body
22h
answered Interaction between a dummy variable and a variable with a quadratic form
1d
comment Using years when calculating linear regression?
Because this is homework I thought giving the answer might be less helpful than hinting how to get to it, as per the self-study tag wiki.
1d
answered Using years when calculating linear regression?
1d
revised Using years when calculating linear regression?
edited tags
1d
reviewed No Action Needed Why use Lasso estimates over OLS estimates on variable subset?
1d
reviewed No Action Needed Is the theory of minimum variance unbiased estimation overemphasized in graduate school?
1d
reviewed Close How to check file integrity of download if checksum is not posted
1d
comment SPSS 16 goes off when i try to create a pie chart
Call IBM, or ask in one of the SPSS forums listed here.
1d
comment Biostatistics and Pure
(1) Hes' talking about Maths posts - Statistics posts barely existed back then. (2) It's more nuanced than a choice between narrowly applicable applied work vs broadly applicable theoretical work. Cox says his theoretical work on conditional inference "was very strongly motivated by practical experience and yet on the other hand, I couldn't say it arose from one particular special type of problem. It arose in a sense from all the applied work I'd done to that point."
1d
comment Unintuitive interpretation of probabilities when doing logistic regression
@statastic: You can use Bayes' Theorem to calculate $\Pr(\mathrm{Class1} |X=x)$ & compare that to fitted models: dnorm(score)*(1/3) / (dnorm(score)*(1/3) + (dnorm(score,10)/2 + dnorm(score)/2)*(2/3)) I think
1d
comment Unintuitive interpretation of probabilities when doing logistic regression
@statastic: ... or is then there something off with your model?
1d
comment Multinomial logistic regression does not match actual data
@Phil: That's what I thought it might be - the intercepts are all huge on the logit scale. (Long experience of making similar mistakes, not smartness.)
1d
revised Treating predictors as numerical or categorical variable in regression
added 329 characters in body