In Paralympic field events, up to 30 classes contest what for able-bodied athletes is a single event (discus, javelin, shot etc.). If each class were to contest each event separately the competition would be overly long; also, the standard of competition may be lower than desirable. Consequently some classes are combined to reduce the number of separate events, and to ensure maintenance of appropriate standards. For intra-class comparison, it is necessary to define appropriate standards for each class. We show that for most classes valid comparisons are possible; but in some cases, they are not. We use this analysis to investigate the "fairness" of the current International Paralympic Committee standards, and show that some of the proposed multi-class combinations are inappropriate. Methods: A table of event results (including all trials for preliminary, semi-final and final rounds) was constructed for Paralympic and World Championship events from 1998 to 2011. Mixed modelling regression was used to attempt to minimise the influence of known effects: gender, year; implement weight; class; and competition type; as well as second order interactions. Provided that selected performances are all sufficiently large, and that one particular condition is met, residuals are expected to follow the two-parameter extreme value (EV) distribution asymptotically. The critical condition is that the distribution of residuals so determined is stationary (invariant in time). Failure to meet this condition is indicated by failing to reproduce a straight line on a log-linear plot of ordered residuals versus magnitude. Given this condition, performances can be compared, even though they come from different classes (or even different events), via percentiles of their appropriate distributions, subject to the uncertainties in the slope and intercept of their respective linear ordered-residual plots. Results: For most event/class combinations, the method proved to be valid, and useful when the total number of performances exceeded 150. However, for several classes (e.g. cerebral palsy classes below F37) it appears that the above analysis of performances does not lead to stationary distributions, and these classes cannot be justifiably combined. Discussion: Reasons for the observed non-stationarity remain speculative at this stage; but it can be observed that significant improvements have been made recently by a small number of competitors in these classes. Relatively low numbers of reported results mean that such a competitor was able to significantly influence the distribution, resulting in non-stationarity. Similar results have been seen in able-bodied events where there is likely to have been a decline in the use of performance-enhancing substances (due to more stringent rules and enforcement). We conclude that the current International Paralympic Committee rules for combined class competition are unfair in a small number of cases; adoption of our methodology should lead to increasingly valid comparisons in all classes as more results accrue.
© Copyright 2012 17th Annual Congress of the European College of Sport Science (ECSS), Bruges, 4. -7. July 2012. Veröffentlicht von Vrije Universiteit Brussel. Alle Rechte vorbehalten.