My blog has been a little bit neglected lately. As you can tell from my recent posts, I've been more interested in getting ready to become a dad than think deep thoughts about statistics. However, I have had some fun with the overlap of statistics and parenting, and that's especially true when it came to trying to choose a name for our baby. I built a Shiny app that implements a MaxDiff exercise that ranks potential names by preference for both you and your spouse (source code here).
MaxDiff is a market research methodology we used extensively during my time at Qualtrics. I hope to have some time during parental leave to dive into the statistical underpinnings for MaxDiff, but for now just know that it takes the difficult task of ranking a large number of items easier by making comparisons between a few of the items at a time. It helps market researchers measure preferences for items and how those preferences differ by group. My team joked all the time about using MaxDiff in normal life decision making, especially around choosing baby names, so I finally got around to making that a reality! A few notes about the app:
I only have 25 hours of compute time per month on my free tier shinyapps.io account, so if it doesn't load that's probably why.
The hierarchical Bayesian model it fits to the data is overkill when you only have two respondents, but it becomes really valuable when you have several hundred.
My wife steadfastly refused to take this app at all seriously, suggesting I'd "use it against her" if I liked the results and she didn't.
The 'recommended name' comes from the highest average preference, even if that is largely driven by one passionate fan, and the other spouse is ambivalent. You can decide for yourself whether that's a good way to come to an agreement.
The random names takes a random sample of the names, where the probability of selection is proportional to how popular the name was according to the 2017 data included in the babynames package in R. The weighting ensures you get a mix of popular and not popular names.
תגובות