Use of the Kalman measurement update theorem for sequential orbit determination incurs the requirement, from fundamental hypothesis, that post-fit measurement residual ratios have a standard normal distribution. When the residual sample size is small, comparison to the normal density curve is not useful because small-sample normal ensembles do not resemble the infinite-ensemble normal curve (bell curve). The quantile-quantile (QQ) plot with Royston-Mi-chael acceptance boundaries for normal distributions is useful for small and large ensembles. Herein I demonstrate the application of QQ-Plots to simulated measurement residual ratios. On real data, QQ-Plots will be used to demonstrate, or deny, that the requirement is satisfied.
展开▼