Reputation
Next tag badge:
98/100 score
26/20 answers
Badges
6 27 63
Newest
 Sheriff
Impact
~533k people reached

1d
comment Reference to Statistical Control Quality
There is a supplement available: wiley.com/college/bcs/0471316482/add/s_index.html
1d
comment Logistic regression coefficient too high - cannot interpret odds ratio
See hauck-donner-effect.
1d
comment Test for one coefficient doubled another one?
(+1) Perhaps worth noting that the likelihood-ratio test may be preferable for generalized linear models: compare the model having $b_1x_1 + b_2x_2$ with the restricted one having $b(x_1+2x_2)$.
1d
comment Performance of Logistic Regression with time
Interesting question. It would help to give a little more detail about the predictors involved. Some may be extensive, some intensive with respect to time (Intensive and extensive properties). If I understand right, the issue with, say, data usage is that an aggregate measure over a short interval is subject to fluctuations that convey little or no information about someone's suitability for the credit product, whereas an aggregate measure over a long interval includes information that may no longer be relevant.
2d
comment A psychology journal banned p-values and confidence intervals; is it indeed wise to stop using them?
Perhaps one way of reading that is as saying the initial significance test that suggested a substance stimulates a certain physiological response is no longer relevant by the time you're publishing your investigations into the effects of different kinds of inhibitors on the dose-response curve.
2d
comment A psychology journal banned p-values and confidence intervals; is it indeed wise to stop using them?
If that's so, what's the answer? Shun inferential statistics?
Apr
26
comment With multinomial regression, how to predict an event and get the ROC curve?
See also How to plot ROC curves in multiclass classification?.
Apr
26
comment interpretation of main effect and interaction
Repeat the plot specified above for a few different values of FINAR. Or use some type of 3-dimensional plot - a contour or a wireframe plot.
Apr
26
revised Is it normal for logistic regression, to have predictors which have good Wald's Chi Sq, but still bad performance?
fixed typo
Apr
26
comment A psychology journal banned p-values and confidence intervals; is it indeed wise to stop using them?
If we accept the premise, the obvious retort is that physicians, psychologists, &c. should consult statisticians when they're having difficulty with statistical analysis - just as statisticians consult physicians, psychologists, &c. when they need to.
Apr
26
comment interpretation of main effect and interaction
See interaction of categorical and continuous variables. All you need do to examine the relationships predicted by your model is plot ETR vs EqRATIO for each of SME, MNE, & LARGE DOM, holding the rest fixed. $\mathrm{ETR}_\mathrm{SME} = 0.135 - 0.030 \times \mathrm{EqRATIO} + \mathrm{constant}$, $\mathrm{ETR}_\mathrm{MNE} = (0.135 + 0.056) + (- 0.030 + 0.001) \times \mathrm{EqRATIO} + \mathrm{constant}$ & so on. Vary EqRATIO over values of interest, & it may aid interpretation to set other variables equal to their means, or typical values.
Apr
26
comment How do you write down a Firth logistic regression model?
The model's the same - just the coefficient estimates change.
Apr
25
comment predict.glmnet() gives same predictions for type = “link” and “response” using family = “binomial”
See predict.lognet
Apr
25
comment R's algorithm for finding glm estimates
See also stackoverflow.com/questions/19226816/… But there's so much fiddling around with error checks & such that @mdewy's suggestion is better if you want to know how the fundamental algorithm works.
Apr
25
comment SPC charts in R - “Error in plot.window(…) : need finite 'ylim' values”
Might be suitable for Stack Overflow if you include a minimal, complete, verifiable example (see ?dput for how to provide your data). I looked at the qcc manual, & there's nothing obviously wrong with what you describe yourself doing - so the devil must be in the detail.
Apr
25
comment Is it possible to do a time series analysis with more than one explanatory variable?
@RichardHardy: The original version, prior to edits, contained only the 2nd question. I've re-opened the question, though it's perhaps rather broad even when focused on regression: a good answer should IMO include helpful references for someone new to time series analysis.
Apr
25
comment Is it possible to do a time series analysis with more than one explanatory variable?
See Hyndman & Athanasopoulos, Forecasting: principles and practice Ch. 9.1 for an introduction to time-series regression with examples using R.
Apr
25
comment How to find standard error of y-intercpet
I've rolled this back to the first edit - substituting "dependent variable" for "intercept" makes no sense.
Apr
25
revised How to find standard error of y-intercpet
rolled back to a previous revision
Apr
21
revised I need a nonsense detector on my model
I hope "nonsense" is a close enough synonym.