Our World Statistics Day conversations have been a great reminder of how much statistics can inform our lives. Do you have an example of how statistics has made a difference in your life? Share your story with the Community!
Choose Language Hide Translation Bar
Staff
How to get a toddler to rank her favorite Easter egg colors

How do you get a 2-year-old to rank nine different Easter egg colors? (Photos by Caroll Co)I’ve blogged before about experimenting with achieving easier-to-peel hard-boiled eggs. My wife and I were able to put this in practice recently while dyeing Easter eggs with our 2-year-old daughter. We found ourselves with nine different dyes to choose from.

I thought it would be interesting to see our daughter’s likes/dislikes, but to try and get a ranking of nine different colors of eggs seemed like an impossible task. Showing her a reduced group of a few at a time seemed doable (such as we did previously in the JMP group in comparing potato chips), and we have some new features in JMP 13 to make this easier.

The MaxDiff platform allows analysis of experiments in which subjects choose their favorite and least favorite items in each choice set, and a MaxDiff designer to create MaxDiff designs. With such a design, I can show my toddler three eggs at a time, getting her to pick her favorite and least favorite each time.

### Designing the study

I’ve blogged before about creating a MaxDiff design using the Custom Designer. In the latest version of JMP, we have a new platform specifically to construct a MaxDiff design. A key differentiation is that the new platform does a better job of ensuring that each pair of treatments occurs in a choice set together. The new design platform is under the DOE menu: DOE -> Consumer Studies -> MaxDiff Design. The MaxDiff design platform takes in one column from a data table representing the treatments under consideration. In this case, it’s the nine different color choices, so we can create a data table with a row for each of the colors:

Choosing the Color as the “X, Factor” on the launch screen, we have a couple of design options to select before making the design:

• Number of Profiles per Choice Set: How many eggs are we going to compare at one time? We’ll stick with three for this.
• Number of Choice Sets: In our case, how many different groups of 3? The more sets we have, the more information, but I didn’t want to exhaust my daughter. I ultimately chose 12. This is a few more than I had originally planned, but I happen to know 12 has the special property that every pair of colors end up being compared (exactly once) in a choice set. For those readers who care about creating these so-called balanced incomplete block designs, we’re looking at making this a lot easier in an upcoming JMP release.

### Collecting the data

The data collection went smoother than we anticipated! To be honest, we didn’t ask a 2-year-old for the best and worst for each choice set of three eggs. We first asked for her favorite of the three, put the favorite out of sight, and got her to pick her next favorite.

The first few rows of the data table from the MaxDiff designer are presented below. The Choice Set column refers to a grouping of three colors presented together. In the Choice column, 1 represents the favorite from the choice set, and -1 is the least favorite.

### Analyzing the data

The new MaxDiff analysis platform can be found under Analyze->Consumer Research->MaxDiff. Selecting our data table, we now have to set the columns to the appropriate roles. Subject ID is the column Subject (although we only have one subject), and Choice Set ID is the Choice Set column. The Response Indicator, where we recorded the selections, is the Choice column. We only have one profile effect – Color, containing the nine different colors (treatments). This is what the platform looks like before running the model:

Looking at the results, we can see the relative ranking of the colors:

I turned the results into a data table, and added a label for the Color column and row colors to look at the results in Graph Builder:

The marginal probability reflects the estimated probabilities if everything were presented together. What we see is that my daughter really liked red, purple, and dark blue, with yellow a distant fourth, followed by the remaining colors.

I found it curious that the three colors were grouped together at the top. Knowing that each pair of colors was compared once in choice sets, I went to examine those particular choice sets. In choice set 2, red beat dark blue. In choice set 3, dark blue beat purple. In choice set 6, where red and purple are together, you might think that red beat purple. But, purple was chosen above red, resulting in that grouping at the top. I guess the ease of the data collection might have been a bit misleading….

### Final Thoughts

The data collection and experimentation for this example were a lot less time-consuming than my typical home experiments, so I did appreciate the MaxDiff design in this case. Unfortunately, by the time I analyzed the data, the eggs in the top three colors were no longer available to see about breaking the near-tie at the top – they had been made into egg salad already. If I'd still had them, I would have asked my daughter to rank that particular set of three. On the positive side, we’ve discovered our toddler likes egg salad sandwiches enough so that we won't be struggling to eat all the hard-boiled eggs as in previous eggsperiments!

Happy Easter egg hunting!

Article Labels

There are no labels assigned to this post.