Masey O’Neill has been a JMP user since she arrived at AB Vista. She discovered early on that it offered functionality she couldn’t find elsewhere. Then, in 2011, her team was introduced to JMP Pro.
“We were just starting to develop our understanding of linear methods,” she explains, “and our JMP representative said, ‘Perhaps you’d like to look at these nonlinear methods such as decision trees and neural networks.’” They liked them quite a bit.
Then they learned about the model validation features in JMP Pro. “And we thought, ‘Yes, that’s really valuable; that gives us a lot more confidence in our models.’”
Masey O’Neill is able to combine all the experiments AB Vista conducts across the globe into a single database, regardless of trial design. She then uses a partition method, perhaps a decision tree, to analyze that data in confidence.
“If the outcome of interest is the efficacy of one of our particular products, we can use that analysis to say in which situation it works best – what are the primary drivers of when our products work most effectively.”
Decision trees let the team examine interactive factors, Masey O’Neill says, bringing considerably more depth to the research and producing much richer insights.
Uncovering the obscure
Research that Masey O’Neill and her colleagues have conducted on Econase XT, an animal feed enzyme and one of AB Vista’s flagship products, is a good example of where JMP Pro has enhanced insight.
The team compiled results on the enzyme from 85 trials, combined them and analyzed them using JMP Pro, and were able to determine with what other additives it performs best. In essence, it allowed them to conduct a single experiment, examining multiple, interactive factors, and uncover otherwise obscure associations.
“We can now say that when you have a corn-based diet, such as in the US, fat is also particularly important in making sure that product works well,” Masey O’Neill explains. “So this allows us to advise our customers that if you’re going to use corn, then you might also want to consider your fat level.
“And we know this from the predictive modeling we’ve done with JMP. The really important point here is that we wouldn’t have been able to divine that from one single experiment. Had we run one experiment that looked at a corn-based diet versus a wheat-based diet, we wouldn't necessarily have been able to look at the fat at the same time.”
Econase XT improves whole-diet digestibility. It enhances amino acid, fat and carbohydrate digestion. More nutrients mean reduced feed costs and less output of waste into the environment – a more sustainable agricultural solution.
‘Proper and thorough’ analysis
Masey O’Neill uses a number of JMP tools nearly every day (analysis of variance is another favorite), because, she says, “proper and thorough statistical analysis” is critical to her work. JMP Pro provides just that, and with ease. Masey O’Neill underscores the software’s user-friendliness; little training is required. Analyses are repeated quickly and efficiently, a boon to productivity.
She also appreciates the value JMP brings in illustrating her findings.
“We use distributions to display the important points from multiple regressions,” she says. “For example, we’ll use decision trees to define the interactive factors involved in predicting a response variate.”
Attempting to explain this to someone who isn’t familiar with statistics can be difficult, she says. But plotting the distributions of all the variables with the JMP Distribution platform simplifies things considerably.
The Distribution platform offers “a very nice, very clear way to display data back to the customer. It’s a very simple thing to do in JMP.”