Reputation
Next tag badge:
109/100 score
19/20 answers
Badges
1 47 110
Newest
 Nice Answer
Impact
~1.1m people reached

12h
comment Can AICc be used to select GLMM models with highly correlated predictors?
Please read the voluminous notes on this site about the pitfalls of model/variable selection. AIC and AICc are restatements of $P$-values so you are essentially doing now infamous stepwise regression.
14h
revised The role of validation in estimation and hypothesis testing
edited tags
14h
answered The role of validation in estimation and hypothesis testing
19h
comment How LDA, a classification technique, also serves as dimensionality reduction technique like PCA
It is clearest to use the term 'dimensionality reduction' to deal only with unsupervised learning methods, e.g., clustering and redundancy analysis. LDA is strictly supervised learning so would create an overfitting bias were it to be used in the first step of data reduction.
1d
comment Can I use Cox Regression here?
Whoops - sorry about that!
1d
comment Can I use Cox Regression here?
I'm not clear on why you would mention the Student $t$-test in a context with binary or time-to-event $Y$. For a survival analysis (which seldom would use a $t$-test) you use the date of entry, date of event or last follow-up, and subtract the former from the latter to get days until event of loss.
1d
comment How should I check the assumption of linearity to the logit for the continuous independent variables in logistic regression analysis?
It is invalid to build models on the basis of univariable analysis. You are using a variant to forward stepwise regression which is known to cause a host of problems.
1d
comment Logistic regression with partially observed variables
If the observations are cross-sectional and not longitudinal/repeated measures and you have some of the variables not missing for later time periods, you can use multiple imputation.
1d
comment Can I use Cox Regression here?
Thanks for the order. Death is the endpoint, modeled with the Cox model as time until death. From the Cox model you can get hazard ratios but also predict the probability of dying within 3m. By using time until the endpoint you get more statistical information/power (by e.g. treating a death at 3.1m as a bad outcome and by considering a death at 1w worse than a death at 3m) and are able to handle loss to follow-up.
1d
revised How should I check the assumption of linearity to the logit for the continuous independent variables in logistic regression analysis?
edited tags; edited tags
1d
answered How should I check the assumption of linearity to the logit for the continuous independent variables in logistic regression analysis?
1d
comment Can I use Cox Regression here?
There are disadvantages of selecting a binary endpoint, including loss of power and having no way to handle dropouts before 3 months.
2d
comment Statistical test for whether matching is required
Matching, as usually practiced, involves arbitrary decisions and discarding observations, hence it loses power and some objectivity. I would never use matching to deal with the problem you've got.
2d
answered How to calculate the probability of success of a Logistic Regression model with a single continuous predictor?
2d
comment Difference between modelling and testing association
To be explicit, a multivariable statistical model is usually the best way to perform a test of association.
2d
answered bayesglm (arm) versus MCMCpack
2d
comment Model selection, issues of judgement
Lately there is a great emphasis on model selection. I think this is frequently misplaced and as Peter said we need to spend our effort on (single) model specification. Note that use of AIC or BIC is at its heart the same as using $P$-values, and results in distortions of statistical inference and bias in regression coefficient estimates.
2d
comment Difference between modelling and testing association
In addition to what Peter said, you have to consider whether unadjusted tests are appropriate. Often they are not. To do an adjusted (partial) test one usually needs multivariable models. Think of a test of association (or partial association) as a test of flatness in a regression model.
Aug
28
comment Rare event logistic regression bias: minimal example
Good. Still remains to be seen whether the amount of bias is large enough to worry about.
Aug
28
comment Rare event logistic regression bias: minimal example
I'm glad you are working on this and look forward to others' comments. Even if there is a bias, the bias correction might possibly increase the variance enough so as to raise the mean squared error of the estimates.