Aug
20
comment Interpretation of $\mathbf{y}^T(\mathbf{I}-\mathbf{H})\mathbf{y}$ in OLS
Thank you very much. Great explanation.
Aug
19
accepted Interpretation of $\mathbf{y}^T(\mathbf{I}-\mathbf{H})\mathbf{y}$ in OLS
Aug
19
asked Interpretation of $\mathbf{y}^T(\mathbf{I}-\mathbf{H})\mathbf{y}$ in OLS
Aug
5
comment Do you include papers that you contributed to but weren't cited as an author in on your CV?
Related: academia.stackexchange.com/questions/1112/…
Jul
6
comment Comparing Cox models
What do you mean when you say that you want to "[...] compare the parameter estimates obtained by these two different methods"? (But I suspect you meant "different models"?). Could you be more specific?
Jul
3
comment What causes differences in estimates of 95%CI of the mean across subgroups in Stata?
What is causing such behavior is the choice of the degrees of freedom for the t distribution from which the quantiles necessary to the calculation of the Cis are derived. For example, in your first example 26.33333 - invttail(20, .025)*3.122499=19.819911, while in your second example 26.33333 - invttail(58, .025)*3.122499=20.082969
Jun
29
awarded Popular Question
Jun
21
comment How to do Cox regression if we have a variable which violates an assumption?
Do you mean that it violates the proportional hazards assumption? You can look into Cox models with time-dependent (or time-varying) coefficients. For example see here: Heinzl, H., & Kaider, A. (1997). Gaining more flexibility in Cox proportional hazards regression models with cubic spline functions. Computer methods and programs in biomedicine, 54(3), 201-208
Jun
17
comment Listing acknowledgments on CV
See academia.stackexchange.com/questions/1112/…
May
27
comment What is the right way of computing baseline hazard rate
Besides piecewise exponential models (Poisson) with splines to model the baseline hazard suggested by AdamO, an alternative is using Royston-Parmar models (flexible parametric survival models)
Apr
16
comment PH assumption: categorical variable (1 category does not interact with time)
Creating dummy variables beforehand is a possible way to go. In this example, the variable drug has 3 levels (2 dummy variables in the model), but I let only one of the 2 dummy variables (i.e. only one of the drugs, here) to interact with log(time): webuse cancer xi i.drug stcox i.drug, tvc(_Idrug_3) texp(log(_t)) - hope the syntax is clear
Apr
14
comment Is $R^2$ useful or dangerous?
@probabilityislogic Do you have a reference for the relation between $AIC$ and $R^2$?
Mar
25
comment Discrepancy between metafor and weighted lm() standard errors
The PDF manual for Stata's vwls command contains the answer to your question: stata.com/manuals13/rvwls.pdf See Section "Remarks and examples".
Mar
14
comment Can p-value = <0.05 with CI that includes 0 using chi2 in Stata?
Quickly: the standard errors used to calculate the test and the CI are different. Take a look at the PDF of the Stata User's Guide (help prtest).
Mar
1
comment A Book for Multiple Regression and Multivariate analysis
I really like Applied Regression Analysis by Draper and Smith. It covers also topic in which you might not be interested (ridge regression, nonlinear estimation, GLM...), but the parts on linear regression are outstanding, in my opinion. The chapters on the geometry of least squares are probably the best part of this book.
Feb
22
comment Is the likelihood a true function?
Why the bayesian tag?
Feb
16
awarded Yearling
Feb
16
awarded Yearling
Dec
23
comment Required number for logistic regression
If you have say 100 subjects, 20 of whom die and 80 survive, then according to this rule of thumb you should have no more than 2 covariates (predictor variables) in your model. In fact, with 2 covariates the condition is satisfied: $\frac{\min(20,80)}{2} \ge 10$, while with 3 covariates it is not $\frac{\min(20,80)}{3}<10$
Oct
16
awarded Good Question
1 2 3 4 5