I am trying to develop a predictor equation for a test that we run multiple times a day. I have over 600 sets of historical data on past test runs. There are three inputs to the process weight (lbs), release pressure (psi) and test room temperature (although, I believe the room temperature should be negligible since we hold the rest room at a controlled temperature and never varies more than +/- 1-degree F). The response is Speed (mph). How can I set this up in Minitab (or other statistical software) to develop a predictor equation for the Speed based on the inputs.

Thanks.

There aer a variety of ways to come up with a predictive equation, depending on how sophisticated you want to get. I would tend to start simple and see what solution works well enough for you. And I would suggest always looking at the data, not just having the software give you some best-fit numbers.

The simplest would be to make a scatter plot of speed as a function of each of the variables (speed vs weight, speed vs temperature, & speed vs pressure). If any of these are perfect straight-line fits with no scatter (or at least close enough got your needs), then you are done. You can get the equation for the best-fit line (even Excel can do this easily) and that is your predictive equation.

If the data on any of the graphs are a scatter-free curve but NOT a straight line, then you ought to look for some other equation to fit the curve - perhaps a parabola or an exponential curve.

If none of these simple graphs is a good fit to the data, then you could try a multiple regression. This can include all three variables in the predictive equation. The R-square value will tell you how good the fit is. R-square = 1 is a perfect fit; 0 means there is no correlation.

This should at least get you started. Let us know how it is going.


Tim