Estimating components of propagated variance in growth simulation model projections

Publisher Website
Google Scholar
Abstract
First-order Taylor series variance estimation equations were embedded in a growth simulation model to estimate propagated variances during growth and yield projections. Variance equations estimated three error components: covariances propagated through predictor variables, covariances from estimated regressor coefficients, and covariances between regressor coefficients and variables. A separate Monte Carlo process was used to estimate the total variance in projected variables caused by simultaneous perturbations in values of initialization variables and in regressor coefficients. Variances estimated by these two procedures were compared over five consecutive projection periods for six variables in a forest growth simulation model. While results agreed closely for the variance in mean stand diameter, disparities increased for other variables later in the model estimation sequence. Disparities were attributed to differences between the populations used in both variance estimation procedures and to possible violations of Taylor series assumptions in the variance estimation equations.

This publication has 0 references indexed in Scilit: