09-09-2016 03:04 AM
All of this is for SAS 9.4M3
We are using dynamic time warping in Proc Similarity to help us find features in depth logs from oil wells. This seems to be working well. The way the method is built you will get different results based on your starting point. The data sets left form the procedure especialy the outpath can with some progarmming be used to find features from training data. Howvere debugging and tuning needs the plot attached. This plot work well from the procedure whne the data points do not exceed 100. In our case we need to test different lengths of the series and being able to make subsetted portions of this would help. As the method give different results with different starting points you have to be able recreate the graph and subset to different sections. Do anybody have an idea?
Also the procedure is meant for time. Here I'm using depth in feet or meters. This needs real numbers. The ID statement only accepts time or integers. If you use integers you have to id these as seconds. This really destroys the automatic graphs. The _timeid_ , _tarpth_ and _inpppth_ variables are really cumbersome to work with. Does anybody have a good trick for using other id's than time and allowing for real numbers?