Distributed lags and truncated Pascal distribution
Hans-Edi Loef
The lag procedures of Solow, var Alphen and Merkies and Maddala and Rao based on the Pascal distribution with a full range of the lag weights from zero to infinity can not detect the relevant (optimal) lag duration. Allowing truncation of the Pascal distribution and using the search process of van Alphen and Merkies leads to optimal values for A and r (the two parameters of the Pascal distribution) as well as the optimal lag duration (range of the lag weights or optimal truncation line). Setting the time lag duration equal to the number of observations leads to the normal search process of van Alphen and Merkies.