Question:  What methods are available to import historical data from website?
Answer:  Three methods are given for importing data from file or spreadsheed formats.  Two types of ASCII file formats are supported being Comma Separated Variable (CSV) and generic tab delimited text fields.  These file types are generally the format provided by websites suppling tabular data.  Cut and Paste operations between any common Windows application is avalilable through the system clipboard using the grid as the entry tool.  Shiraz native file format is not compatible with other applications or spreadsheets.

Question:  Why do the estimated value/sign and all other statistics have zero values for the first few rows of data elements?
Answer:  The estimated and target values are only valid beyond the first batch of input points whose length is defined by the number of memory taps.  For example, if the number of memory taps is setup for 10 then the first 10 datapoints correspond to the 10th output.  Preceeding target values will not have a complete set of inputs defined for operation due to being less then the number of memory taps.

Question:  Why is the number of output terms a read-only field preset to 1?
Answer:  Currently the predictor only can have a single target or output term active which prevents extrapolation with multiple inputs.  In order to extrapolate the prediction must propogate the value for each input which implies one output to one input pairing.  Future editions of the program will enable the usage of multiple outputs purely for allowing multi-input extrapolation operation.

Question:  Why is the number of input terms a read-only field?
Answer:  The number of input terms is a product of number of input columns and memory tap depth.  There would be no use for manual overriding of this field since then previously mentioned relation is always true.

Question: Why select between 1 pass or 2 pass estimations?
Answer: The 1 pass estimator provides exceptional performance but suffers from long term trend effects larger then the active number of samples window.  For example, it you had a data set with a gradual long term slope for 100 points and the estimator window size was set to 15 taps then the longer length slope would result in a bias error.  The 2 pass estimator correctly models long term effects greater then the window size.  Care must be used when selecting the window size for the 2 pass estimator to be larger then rapid short term effects such as a rapid slope transition.

Question: Why are the extrapolated values different if training period is "Continued" a different number of times?
Answer: The Trendfinder training method uses a Monte-Carlo algorthm which terminates depending on the relative statistical errors between the Training and Testing vector sections.  This method is used to eliminate the "Over Training" condition which serves to limit Trendfinders ability to estimate semi-random sequences.  If the training is allowed to terminate with similar conditions then the extrapolated values will be exact.  Continued training, while useful for first time model developing, should not be used during the extrapolation operation.