Sawtooth Software: The Survey Software of Choice

A Discrete Choice Take on Halloween: Results

As promised in my fun Halloween article, we have the results!  With a significantly smaller network than @fivethirtyeight, our sample size is on the lower end at n=73. That being said - MaxDiff does an amazing job with small sample sizes and the data proves it!  Comparing our n=73 results to the n=8,300+ of @fivethirtyeight, we arrive at the same top 5 candies. Now we just have to argue over who is right!

Read More

MaxDiff Simulator, TURF Analysis, and Export to PDF now available in Discover

Our team has been hard at work making Discover better. Discover now has a MaxDiff Simulator, a powerful TURF analysis tool for MaxDiff, and the ability to export your survey to PDF. 

Read More

With HB, the Goal Is NOT to Maximize Individual-Level Fit (RLH)

Hierarchical Bayes (HB) estimates high-quality individual-level utilities for CBC and MaxDiff (Best-Worst Scaling) despite sparse data for each individual. HB’s goal is not to maximize the fit to the individual-level choices (the RLH statistic, or Root Likelihood). This could lead to overfitting and the utilities may do poorly in predicting new choices (such as holdouts or real market choices) outside the scope of the data used in utility estimation.

Read More

A Discrete Choice Take on Holiday Movies

Research shows that on Christmas day, holiday-themed programming spikes as households curl up in front of the television for a dose of holiday cheer. Timeless classics like “How the Grinch Stole Christmas” and “It’s a Wonderful Life” come to my mind. And thanks to @RottenTomatoes, we have a list of the 50 Best Christmas Movies of All Time.

Now the last thing you want to do on the holidays is argue with family. So how will you decide which Christmas movie to watch on December 25th? Why, MaxDiff of course!

Read More

On the Effect of None Position for CBC Studies

In December 2018 we fielded a CBC study using Lighthouse Studio, collecting 201 respondent records via Amazon’s Mechanical Turk panel.  The subject matter was restaurant choice and we used a screener question to qualify respondents who ate at restaurants at least once per month.  We were interested in whether placing the None concept as the right-hand concept vs. placing it as a concept along the bottom of the task would affect the frequency of None usage.  We found that the placement of the None concept did not lead to a statistically significant difference in the None usage.

Read More