Reputation
26,205
Next tag badge:
93/100 score
33/20 answers
Badges
1 43 106
Newest
 Necromancer
Impact
~840k people reached

16h
comment Insignificant coefficients in Logistic Regression after LASSO variable selection
As @mark999 stated this approach is not statistically valid. You can't pretend that estimates can be unpenalized when you originally used penalized maximum likelihood estimation. The problem with penalized estimates is that you can't get statistical inference unless you are Bayesian.
1d
revised How to rescale “linear predictor” in drawing nomogram with “rms” package in R?
edited tags
1d
answered Is validation set always necessary?
May
21
awarded  Necromancer
May
19
awarded  Nice Answer
May
18
comment ROC curve cut off and weights
Any time you use a lot of $ in R commands there is a much better way. Sensitivity and specificity are in backwards time-order and reverse information flow (condition on the future to predict the past). They should play no role in a prospective probability estimation setting. Concerning risk estimation study binary logistic regression.
May
18
comment ROC curve cut off and weights
Your R code could be significantly improved. I suggest you study R a bit more. And your question is phrased in such a way as to indicate that you do not fully grasp risk or utility/loss/cost functions. This problem should have nothing to do with ROC curves. Risk estimation is what leads to optimum decisions for individual patients.
May
15
comment How to report machine learning research?
Since you are using classification as opposed to estimating probabilities, I don't think this is a particularly good approach.
May
15
awarded  Necromancer
May
13
comment Calculating goodness of fit and choosing the right model (R)
There is k-1 where k is the number of unique Y values. Think of the intercepts as the empirical distribution of Y given X
May
13
awarded  Nice Answer
May
11
awarded  Revival
May
8
comment Why can I use the posterior probability of a classifier as a new classifier?
You can do all this without classification, just stick with probabilities. The logistic regression model is a more direct probability model, though, with fewer assumptions than discrim. analysis. If the assumptions of discrim. analysis hold, the logistic regression assumptions automatically hold.
May
8
revised Calculating goodness of fit and choosing the right model (R)
Expanded to answer the OP's follow-up question
May
7
comment Calculating goodness of fit and choosing the right model (R)
If you are running the latest version of R (and if you are using RStudio you have installed its latest version too) rms and its dependencies work fine.
May
6
revised Calculating goodness of fit and choosing the right model (R)
edited tags
May
6
answered Calculating goodness of fit and choosing the right model (R)
May
3
comment Handling Missing Values During Test Phase
This will be very inefficient if the variable has any predictive value. Multiple imputation is strongly preferred. Here is a reference that studies several approaches to the problem: citeulike.org/user/harrelfe/article/13265778
May
1
comment Finding degree of polynomial in regression analysis
Are you planning on doing any statistical inference (confidence bands, hypothesis tests, etc.)? That would alter the approach.
May
1
answered truncated quantile regression in R