Beßt Regression - Features

Home

Endorsements

Abstract

Features

Examples

Beßt Endorsements

I have been extremely pleased with the Best Excel Add-in. I have used the tool for the last five or so years. The service from Best has been very good whenever I have had any issues as operating systems have evolved. We had a case where our supplier failed the qualification of a new component. The engineer came to me for advice. Fortunately I had previously developed an engineering-based non-linear empirical model of the supplier process using Best and we used the model to make recommendations for process changes that resulted in the qualification to pass on the next run. This saved us significant time at a crucial time of the project.

Ronald Dombro Jr., PhD

Chief Technology Engineer
January 2019

To the Best Regression staff:

I honestly don't remember how I found you, but I'm sure it was via an internet search engine years ago. At least 15 years ago, I'd say. I was trying to find Arrhenius parameters for kinetics data and was terribly frustrated by the poor fits I was getting due to problems of scaling and correlation between A and Ea. I had used expensive proprietary SimuSolve software (from Dow), SAS nonlinear regression (using Marquardt grid search with no luck), and later JMP. I also tried setting up my own regression using Excel's Solver. All bad results. I'd have a set of data that was easily fitted visually with a nice smooth curve, but the regressions would always fit it with an exponential that would climb rapidly and then plateau, going through the data in an average sort of way but not fitting it well at all. I think it was following my frustration with Excel Solver that I may have been randomly searching around on the internet for Solver replacements, and I believe it was in that context that I found Besst.

At the time, it was very difficult to buy software that wasn't on Dow's IT department's "approved" list, and had you been selling it for $100's I never would have bought it (after all, I had no idea if it would work any better than the other approaches I had tried). But, you sold it as a subscription for a modest price - $49 - and so I was able to sneak it in on an expense account.

I was thrilled with the results. I didn't subscribe every year. I think maybe only 3 or 4 years, actually. I did it whenever I was working on a problem that required good nonlinear regression capability, and I didn't have that situation every year.

For what it's worth, my 2 cents would be that your product is so much better than Solver, it would be ideal if you could work with Microsoft and have it included as part of Excel, either as a replacement for Solver or as an adjunct.

At Dow I co-oped in Midland in 1980, and joined Dow in Texas in 1981. I worked entirely in the polymers arena, in R&D; first in Polyolefins Research working on polyethylene additives/stabilizers, then on ethylene elastomer product development (including 9 years at the DuPont Dow Elastomers JV), and then finally in packaging (ethylene copolymers and reactive extrusion). I retired in 2011.

I used Besst several times during my polymer science career at Dow Chemical and DuPont Dow Elastomers. I used it when I was analyzing kinetics data, whether for chemical reactions or rheological properties. Typically, I was fitting some form of the Arrhenius equation, although I used it for other nonlinear equations as well. I had previously attempted solutions using Excel's nonlinear regression algorithms (Solver), SimuSolv (a chemical kinetics software developed by Dow Chemical), and SAS software including JMP with its nonlinear regression module. Despite using Marquardt search algorithms in the latter, none of these was able to converge on reasonable parameter estimates when the parameters were correlated and had scaling issues, as was typically the case with these kinds of datasets and parameters.

When I finally found Besst, I was delighted. It was easy to use and affordable (important because I only used it occasionally since kinetics analysis wasn't a continual feature of my job; I probably bought the annual subscription 4 times over the course of 15 years or so). And best of all, it consistently and quickly converged rapidly on solutions that fit the data very well. I highly recommend Besst for any regression problem, but especially those that defeat popular statistics programs.

What would be fabulous would be if this were a built-in add-on package for Excel, i.e., if you licensed it to Microsoft and they included it with every Excel package as an alternative to Solver, or replacing the engine behind Solver. I suspect it is the kind of tool that many people—not just chemists and engineers but also finance people—could benefit from if they knew about it and it were readily available. Yet, like me, most of these people would probably only have need for it on an occasional basis. It would be great to have it available on an as-needed basis right in the most popular spreadsheet software.

Regards,

Robert Johnston, retired polymer science, engineering, and technology consultant
February, 2019