Candy Preference and Age


Less is more. Stay pure. Stay poor.
I was getting ready to go to bed last night and my wife asked me what type of chocolate candy I prefer from a list of 4, then what type of sugary candy,..., and lastly old timey candy. Then whatever site she was on kick out an estimated age for me:

42 years of age, which was off by some years, but probably close enough

She got a 34, which was closer to her age.

Question, what type of model were they running if they had four categorical questions with four response groups, respectively. This is just a guess that the test is static and not adaptive based on prior question. And it kicks out a numeric DV.

Just a multiple linear regression model based on some survey or database of preferences is my guess. What do you think?
Is it this one? Unfortunately I can't really take it because I don't recognize most of them, not being American.

I might be wrong, but I'm not sure there's a sophisticated model behind it. Many of these kind of quizzes just use arbitrary points for each option.

But maybe I underestimate Buzzfeed.


TS Contributor
my guess would be a simple model based on age of the candy - this should be reasonable enough given that tastes develop in childhood. So, if I could find 4 camdy types that were popular at different times and I can assume that most respondents are in the age bracket of 60-20 I could get a precision of roughly plus minus 5 years.
Does this figure?


Less is more. Stay pure. Stay poor.
I hadn't looked up the instrument, so thanks Jules. Perhaps I will take it then change one selection and see how that changes the output.

I was inititially thinking it was just Bayesian network but that intuitively wouldn't kick out a age number unless they just averaged the ages of people in the subgroups but that would require a lot of data (high dimensionality).