Standard tests for linearity vs STAR effects can be fooled by the presence of outliers. The following demonstrates a test adjusting for outliers using iterated weighted least squares from van Dijk, Franses and Lucas(1999), "Testing for smooth transition nonlinearity in the presence of additive outliers", JBES, vol 17, no 2, 217-235. The example uses the lynx data analyzed in Terasvirta(1994)intentionally contaminated with outliers. The final test is done using @STARTEST with the WEIGHTS option, where the weights are generated by iterating weighted least squares.