Impulse Response Identification and Causality Detection for the Lydia-Pinkham Data

The Lydia-Pinkham data is analysed using a recently developed system identification algorithm. For an observed time series this yields an estimate of the process impulse response which we argue is a more robust modeling device than the traditional autoregressive moving average model for econometric lime series. Our results are compared with some earlier results for the Lydia-Pinkham time series. Further, a new multivariate causality test on the data dramatically reveals a uni-directional casual relationship from log-advertising expenditure to sales