Fig. 8 show that the data are not IID. In effect, the ordered plot suggests a nonstationary distribution with central tendency that follows an exponential learning curve. Fitting an exponential equation to the data using least-squares yields Hours=2.86 Unit−0.33, and is plotted in Fig. 6. We can then look at the deviation of the data from the learning curve to see if the deviations are IID, and if so, what distribution best explains the deviations. Fig. 9 shows the ordered plot of the deviations, Fig. 10 shows the scatter diagram, and Fig. 11 shows the correlation plot. All three indicate that the assumption of IID is likely. Fig. 12 shows the histogram and distribution fit of the deviations. Based on this analysis, the installation of valves of this type could be represented in a model using the formula 2.86Unit−0.33+Beta −0.233,0.194,1.49,1.26, where Unit is a dynamic variable that represents the sequential number of the unit being installed the data for this example was artificially generated using the formula 3Unit−0.34+Uniform−0.2,0.2. In many cases several activities depend on common underlying conditions. As the conditions change, so do the distributions of the activity times. When this happens, the data from the various activities are correlated. In these cases, the underlying conditions need to be identified and modeled, and the distributions used for the different processes expressed as functions of these conditions. It is also possible to determine n-variate distributions to explain the correlated data, but this is not easily done without many further assumptions regarding the underlying distributions and 1:1 correspondence among data sets.