I recently attended a great course by Odd Aalen, Ornulf Borgan, and Hakon Gjessing, based on their book ’Survival and Event History Analysis: a process point of view’. Among the many interesting topics covered was the issue of how to interpret changes in estimated hazard functions, and similarly, changes in hazard ratios comparing two groups of subjects.
Jonathan Bartlett
Leveraging baseline covariates for improved efficiency in randomized controlled trials
In a previous post I talked about the issue of covariate adjustment in randomized controlled trials, and the potential for improving the precision of treatment effect estimates. In this post I’ll look at one of the (fairly) recently developed approaches for improving estimates of marginal treatment effects, based on semiparametric theory.
The Hosmer-Lemeshow goodness of fit test for logistic regression
Before a model is relied upon to draw conclusions or predict future outcomes, we should check, as far as possible, that the model we have assumed is correctly specified. That is, that the data do not conflict with assumptions made by the model. For binary outcomes logistic regression is the most popular modelling approach. In this post we’ll look at the popular, but sometimes criticized, Hosmer-Lemeshow goodness of fit test for logistic regression.