Exponential regression with Type I censoring

Fits exponential regression models using maximum likelihood estimation. Data may be subject to Type I censoring.

Given covariates X [n x p] and target T [n x 1], the function fits an exponential regression model:

T_i ~ Exp(theta_i) , i = 1, ..., n

theta_i = Exp(X*beta)

using maximum likelihood estimation. The covariate matrix X should not include a constant vector. Parameter estimates are obtained using Fisher scoring with each iteration solving a weighted least squares problem. The method allows for type I censoring with a fixed censoring cut-off point c > 0. To analyse censored data, you must pass a vector of censoring indicators delta [n x 1]. The vector delta can be omitted if data is fully observed. When delta = 1, the data point is fully observed; delta = 0 implies a censored data point. Only type I censoring is supported where the maximum follow-up time is the same for all participants.

An example of how to use the function (testfit.m) is included.