Reputation
6,767
Top tag
Next privilege 10,000 Rep.
Access moderator tools
Badges
1 20 37
Impact
~560k people reached

Apr
10
revised How high is too high with Cronbach's alpha?
edited tags
Apr
10
comment How high is too high with Cronbach's alpha?
To me this is a more plausible explanation than the otherwise good ideas about CMB, SD, etc.
Apr
10
revised How high is too high with Cronbach's alpha?
added 15 characters in body
Mar
19
comment Validate precise scale with less precise scale
Using the less-precise scale you have made just a single measurement? If so, your job of validation is going to be very difficult.
Mar
6
revised Interpreting exp(B) in multinomial logistic regression
If it "makes sense" it will hardly be "confusing"
Mar
6
comment With a binary Y, why are R's lowess fits so often flat?
I've found that the plsmo command used with the Hmisc package in R provides an excellent solution. Not only are the smoothed lines more sensible, but the graphs provide for separate lines for subgroups, with tick marks indicating the N-sizes at different levels of X. inside-r.org/packages/cran/hmisc/docs/plsmo
Mar
6
comment Interpreting exp(B) in multinomial logistic regression
Where you say "the new odds ratio will not be 6" I would say "the new odds will not be 6...but the ratio of the new to the old odds will be 6."
Feb
18
awarded  Notable Question
Feb
17
comment Correlation matrix for categorical variables
You could look up 'measures of association.' There are a variety of them that are applicable to categorical data. Correlation per se is suited to continuous data, and depending on the context it is either meaningless or impossible to compute a correlation between, say, ethnicity and geographical region.
Feb
14
comment Modeling rare events
Fraud is 200 times as common in one dataset as in the other. With the 2nd, 180 cases of fraud is not terrific, not terrible as a basis for modeling, logistic or otherwise, but as @Bjorn said, you should account for any differences between the two databases if you want training-set modeling to be validly tested by validation-set results.
Jan
25
comment Why use logistic regression instead of SVM?
Helpful, thanks.
Jan
24
comment Multiple regression - A question about dependent variable
For best learning and to draw the best answers, avoid broad "should I?" and "must I?" questions. Instead, consider what will be the consequences of certain choices, then ask for more info about those consequences if you have specific questions.
Jan
24
comment Why use logistic regression instead of SVM?
Fidgety? Please elaborate.
Jan
17
answered Evaluating predictions consistently
Jan
17
comment How to demonstrate a third quantitative attribute in a scatter plot effectively
3-D scatterplot would preserve granularity.
Jan
17
awarded  Strunk & White
Jan
17
revised What type of regression for qualitative dependent variable
edited title
Jan
17
comment adding interaction term gives opposite relation
"How to interpret" questions tend to mean "teach me about this topic." Very broad, and likely covered in previous threads on this site. Suppose you read through them and then see whether you still have a specific question about your situation.
Jan
17
revised Is it valid to say higher Beta values are more significant in a regression model
added 7 characters in body; edited title
Jan
13
revised predictive model for mutation data (all binary)
if these edits don't correctly capture your intent, please roll back.