statistical model

Dining out with Bayes Theorem (Updated)

Imagine you’re looking for a place to eat. Say for example you’re in the mood for Mexican food. So, you go to Yelp and type “Mexican” into the search bar and up come a variety of results.

After scrolling through the list, you pick one that seems like it could be good. It’s nearby, not too expensive, and most importantly it gets a positive rating on the basis of a fairly decent number of reviews. Let’s say it gets 4 stars based on 17 reviews.

Here’s the question I pose to you – just how much can you trust this 4 star rating? How do we know that this rating is a fairly accurate reflection of the quality of the food and overall dining experience at this restaurant? And how do we know that this rating is not merely due to some other extraneous circumstance? (more…)

Coming to grips with failure: A strength in science and in restaurant rankings

We try to be as scientific in our thinking as possible here, and a major strength in any field of science is a willingness to accept that methodologies are rarely if ever perfected – this combined with a drive and determination to continue to attempt to make improvements in these methodologies anyway. All for the sake of advancing, however minutely, our understanding of the world.

So, after we ran into some problems during our first attempt to come up with a definitive list of the best places in and around Charlotte to get burgers, I decided to make some tweaks and changes to our restaurant model. We can think of the first model as version beta and the current model as version 1.0, with the explicit understanding that there will likely be many future versions to come. (more…)