The data set medfcoh.dat consists of lifetables for 46 cohorts of female medflies (Ceratitis capitata). For each cohort, the number of flies alive at the beginning of each day was recorded. The cohorts consisted of about 4000 flies each.

These data were used to illustrate a stochastic model for samples of curves and especially hazard functions and trajectories of mortality which are based on the Karhunen-Loeve expansion of a stochastic process in eigenfunctions and eigenvalues. In addition, this model introduces a time transformation in dependency on covariates and associated variables. While such time-transformation models are mainly of interest in survival applications, they may also be applicable to other situations where one collects a sample of response curves.

In the example application based on this data set, a nonlinear function of the mean lifetime of cohort can be estimated and applied to implement a nonlinear time transformation for each cohort, see Capra & Müller (1997).

The data were collected by Professor Carey of UC Davis and collaborators in a medfly rearing facility in Mexico; details of the data collection can be found in Carey et al (1992). See also Carey (1993). The scientific question here is to describe and model the trajectories of mortality for a sample of cohorts. Additional statistical problems revolve around the estimation of a hazard function from lifetables (see Müller, Wang and Capra, 1997; Wang, Müller and Capra, 1998).


Müller, H.G., Wang, J.L., Capra, W.B. (1997). From lifetables to hazard rates: The transformation approach. Biometrika 84, 881-892.

Wang, J.L., Müller, H.G., Capra, W.B. (1998). Analysis of oldest-old mortality: Lifetables revisited. Ann. Statist. 26, 126-163.

Data format:

Each cohort has 174 lines;

First line: Cage number

Lines 2-174: day (starting with day 0), followed by number of flies alive.