Calculating and testing sequential variables from EPG datasets using SAS

One way to study hemipteran feeding behavior is by measuring voltage changes in an electrical circuit that includes the insect and plant, via an Electrical Penetration Graph (EPG) monitor. This procedure results in a recording of waveforms that represent stylet penetration events through time. One approach to analysis of these data is to select specific waveforms and calculate averages of their durations or frequencies, per event, per probe, or per individual insect. SAS statistical analysis programs have been developed to make such calculations. However, this approach ignores information contained in the often stereotypical, sequential ordering of waveform events. We show how SAS can be used to extract, average, and statistically test among biologically sequential variables such as "time to first phloem penetration" from standard EPG data files. We demonstrate the process using data from aphids and psyllids, and discuss both the advantages and limitations of using these new response variables for EPG analysis. The main advantage is to increase the opportunities to discern significant differences among treatments in EPG research. The primary limitation is the need to guard against statistical abuse because one data set can now give rise to hundreds of variables.