Supplement 1. The Splus code for estimating parameters and calculating the posterior probability distributions for DA risk metrics.

File List
S-Plus code:

ParameterEst.ssc

kalman_ests.m

PostPDFPlots.ssc

All files are in ASCII text.

Description The file ParameterEst.ssc contains Splus code to estimate m
and s2 using ML, running sums, and slope
methods. The ML method allows missing data; the running sum and slope methods
do not. For these methods, missing data points should be replaced with interpolated
values. I use simple linear interpolation. Basic diagnostics are done also:
normality, trend and autocorrelation tests. Outliers are tested for and removed
using DFFITS criteria.

The file kalman_ests.m contains matlab code to estimate m
and s2 using the Kalman filter algorithm.

The file PostPDFPlots.ssc calculates posterior probability distributions
for the parameters, m, s2, and s2np,
and the risk metrics, l, probability of eventual
90% decline, and expected total probability of 90% decline within given time
frames. This code implements the algorithm given in the section Posterior
probability distributions in the text and produces Figs. 8 and 9.