Publications by Joel Cadwell

The MaxDiff Killer: Rank-Ordered Logit Models

27.02.2013

Compared to MaxDiff (Sawtooth Software), ranked-order logit modeling:simplifies data collection without needing additional software to generate experimental designsreduces respondent burden making the task easier and seemingly fastercollects more data from individuals (complete ranking data for all alternatives) eliminates the ...

9276 sym 1 tbl

Counts may be ratio, but not importance

01.03.2013

I can see from those of you who have contacted me that there is still some confusion about the claims made by Sawtooth that MaxDiff estimates can be converted to ratio-scale probabilities.  Many of you seem to believe that if attribute A has a MaxDiff score of 20 and attribute B has a MaxDiff score of 10, then A must be twice as i...

2405 sym 1 tbl

Let’s Do Some Hierarchical Bayes Choice Modeling in R!

06.03.2013

It can be difficult to work your way through hierarchical Bayes choice modeling.  There is just too much new to learn.  If nothing else, one gets lost in all ways that choice data can be collected and analyzed.  Then there is all this output from the estimation procedure that you have probably never seen before.  What is this Markov Chain Mon...

4624 sym R (4481 sym/5 pcs) 2 img 5 tbl

Does It Make Sense to Segment Using Individual Estimates from a Hierarchical Bayes Choice Model?

24.03.2013

I raise this question because we see calls for running segmentation with individual estimates from hierarchical Bayes choice models without any mention of the possible complications that might accompany such an approach.  Actually, all the calls seem to be from those using MaxDiff to analyze the data from incomplete block designs.  For examp...

8198 sym R (1687 sym/2 pcs) 2 img

Got Data from People? Take Dan Ariely’s Coursera course.

26.03.2013

A Beginner’s Guide to Irrational Behavior started yesterday.  One might not immediately think that such a course would be relevant for statistical modeling.  Well, it is if your statistical modeling uses people as informants.  If the data come from individuals responding to inquiries, then it would help if we understood the response genera...

2049 sym 2 img

Halo Effects vs. Intention-Laden Ratings: Separating Baby and Bathwater

08.04.2013

Are halo effects real or illusory?  Much has been written arguing that rating scales contain extensive amounts of measurement bias.  Some tells us to avoid ratings altogether (What do customers really want?).  Others warn against the use of ratings scales without major adjustments (e.g., overcoming scale usage heterogeneity with the R pac...

19444 sym R (1015 sym/1 pcs) 4 img 4 tbl

A Call for Context-Aware Measurement

25.04.2013

Context awareness seems to be everywhere, and everyone seems to be saying that context matters.  Gartner foresees “a game-changing opportunity” in what it calls context-aware computing.  The title of their report states it best, “Context Shapes Demand at Moments of Truth.”  Their reasoning is straightforward.  They assume that what yo...

15929 sym 4 img

Incomplete Data by Design: Bringing Machine Learning to Marketing Research

06.05.2013

Survey research deals with the problem of question wording by always asking the same question.  Thus, the Gallup Daily Tracking is filled with examples of moving averages for the exact same question asked precisely the same way every day.  The concern is that even small changes in the wording can have an impact on how people percei...

12412 sym 4 img

Why doesn’t R have a MaxDiff package?

28.05.2013

Almost once every year someone asks if R has a package for running the MaxDiff procedure sold by Sawtooth.  One such inquiry recently received a reply with a link showing in some detail the R code needed to generate a balanced incomplete block design, input the best-worst choice data, and use the mlogit R package to estimate the pa...

4636 sym

The Reorderable Data Matrix and the Promise of Pattern Discovery

12.06.2013

We typically start with the data matrix, a rectangular array of rows and columns.  If we type its name on the R command line, it will show itself.  But the data matrix is hard to read, even when there are not many rows or columns.  The heat map is a visual alternative.  All you need is the R function heatmap( ) from stats, and the...

8491 sym 8 img