APPLICATION OF HIGH-BREAKDOWN ROBUST REGRESSION TO TUNED STOCK ASSESSMENT MODELS

Many tuned assessment models, such as sequential population analysis a nd nonequilibrium production models, are cast in the form of least-squ ares minimization routines. It is well known that outliers can substan tially alter the results of least-squares methods. Indeed, in the proc ess of conducting stock assessments, much time and effort are often sp ent in discussing the merits of individual data points and in evaluati ng the impact that including or excluding them has on the perceived st ock status. Unfortunately, straight-forward statistical tests for dete cting outliers have been developed only for univariate statistics or f or the simplest of linear models and are generally useful to test for a single outlier only. In this paper, we apply a high-breakdown robust regression technique, least trimmed squares, to two assessment models using North Atlantic swordfish and West Atlantic bluefin tuna as exam ples. We illustrate how robust regression can be used as an initial st ep in statistically detecting outliers before the more efficient least -squares minimization can be used.