A kernel as a weighting function was used by Nadaraya and Watson in 1964 to estimate values as a locally weighted average. This produces a smoothed (non-causal) value that can be used to estimate future values.
Estimating a future state given a selective past (lagged variables) is important but provides only a fragment of the information necessary in order to forecast. The predictive distribution on the other hand contains all information about this future state given the selective past.
Recommendations...
- Use it for discretionary estimates only,
- never use (but you may) it in any signaling mode.
- But again, it is "discretionary"... In another context, you can only blame yourself, and not the creator ; )=
UPDATES:
- Band calculation got optimized for optimal probabilities.
...And the same goes with VP-R tools & Regression channel bands.
- New templates & default parameters
To download the current release... Click this >>> download/file.php?id=3449882
For more info about this system... Click this >>> viewtopic.php?p=1295387306#p1295387306
