The Application of Likelihood Ratio Test to Blast Vibration Analysis

The dedicated documenting of blast design and vibration monitoring results forms part of the best practice methodology for today’s blasting operators. This allows regression curves to be constantly updated and in turn compatible charge weight determinations are based upon the most current data. The point can arise however when so much data becomes agglomerated that variations in blast design, geology or monitoring location response and their effects can pass unobserved and start to have an adverse effect on the forecasting capability of the regression due to a deterioration of the standard error statistic. The simple solution would be to only compile like-for-like blasting returns, but this prospect can quickly become unworkable due to the number of potential variables. In addition it is very difficult to spot subtle variations by simply examining a standard scaled-distance curve mainly due to the intrinsic data scatter that is always present together with the logarithmic nature of the regression curve. If indeed a variation is thought to be apparent, how can the degree and magnitude of its influence to be quantified? The Likelihood Ratio Test when applied to regression data can diagnose subtle variations between data sub-sets and provides a comparative statistic as to the degree of variation. By analysing datasets in such a way, levels of compatibility are obtained, indicating which sets can be combined in order to increase prediction capability.