"Rad" wrote in message <email@example.com>... > Hello, I?m trying to determine the feedback and input delays for a 3-input 1-output NARX network using target-target auto-correlation and input/target cross-correlation, respectively. > > For the target-target autocorrelation, the max. lag is at 0, and the value of the auto-correlation coefficients decreases as the number of lags increase; therefore, I?m trying to determine the significant target-target auto-correlation lags by correlating the target with a random series and finding the 95% significance threshold ? thank you Greg for this recommendation.
ALL values above the threshold are significant. Start with the smallest significant lags and only use as many as you need.
> I tried this using a simple dataset (simpleseries_dataset, obtained using help nndatasets; 1- input 1-output, 100 observation) and it works fine. > However, when I try the same approach with my data (attached to this post as target_data.xls; 3 inputs, 1 output, 32063 observations), I end up with a significant lag of 7810, which appears quite large to me.
Wrong approach. Find ALL significant input and feedback lags. Choose the smallest subset of the smallest significant lags that will yield acceptable results..
> If this is not the correct result, how can I determine the proper 95% significant threshold?
Your threshold calculation is probably ok. However, ALL correlations with absolute values greater than the threshold are significant.