Michelin 2015

[oakapple]

Michelin stars and Yelp stars are correlated, and the correlation is statistically significant.

[Sneakeater]

So now we know who the famous inspectors are.

[tighe]

That analysis is such a methodological mess and the interpretation of the results so vacuous, all it really proves is that Silver should stick with calculating averages of other peoples' surveys.

[oakapple]

That analysis is such a methodological mess and the interpretation of the results so vacuous, all it really proves is that Silver should stick with calculating averages of other peoples' surveys.

 

I assume we'll be seeing your superior analysis in short order?

[tighe]

"Extremely statistically significant" is the equivalent of saying someone is 'extremely dead' and a sure red flag that the analyst doesn't actually know what they're doing from a methodological point of view. He models ordinal variables using a method that relies the assumption of the variables being continuous, and even if you close your eyes and wish away that issue the results show that the relationship between the two ratings is weak, at best. I could go on....

 

No, I will not be re-doing the analysis since the value is trivial.

[Robert Brown]

Oakapple, I wouldn't call taking the opening of my post out of context rotten, but it has nothing to do with Pete Wells or the Academy Awards. I'm talking about the guide itself and making the obvious point, when viewed in its entirety, that it is a big waste of time and fails completely as a guidebook. The star ratings are no different than hype words that try to say everything and end up saying nothing, while the text is a bunch of indisciminate or undiscriminating babble that unlike Trap Advisor or Zagat doesn't even allow for reading between the lines. It's at best a bare-bones directory less inclusive than those two other publications. Whether you call the restaurants terrible or, in the case of the Michelin, mediocre or "not worth the voyage", it doesn't really matter, practically speaking.

 

The title escapes me, but about ten years ago a former Michelin inspector wrote an exposé about how corrupt the inspection system was and how the inspectors often didn't visit certain restaurants, all of which begs the question as to who these inspectors are, what are their qualifications and past experiences and how they go about their job. I strongly suspect the Michelin people won't be truly forthcoming about this. Surely the inspectors don't write the descriptions since all of them read the same. Also what do ten-year old NYT restaurant reviews have to do with anything? The end result of all this is what I have always believed: first restaurants present a bad form, or are unsuited, to be reviewed, and that the only person you should rely on is yourself and the research that you do. You can tell a lot more about a restaurant from studying its website. It's where I go first because I get the best indication available as to whether it fits what I look for in a restaurant and, most important, the aspects of contemporary restaurant dining that I can't abide.

[oakapple]

"Extremely statistically significant" is the equivalent of saying someone is 'extremely dead' and a sure red flag that the analyst doesn't actually know what they're doing from a methodological point of view. He models ordinal variables using a method that relies the assumption of the variables being continuous, and even if you close your eyes and wish away that issue the results show that the relationship between the two ratings is weak, at best. I could go on....

 

No, I will not be re-doing the analysis since the value is trivial.

 

If you think all Silver does is "calculate the average of other people's surveys," I don't have much confidence that you've read his work.

[Adrian]

FiveThirtyEight is terrible and is ruining the goodwill Silver built up during his stint with the NYT. That piece is a good illustration why.

[tighe]

By his own description, Silver's electoral forecasting is based on a meta-analysis of various published surveys. Yes, he applies different weights to each survey depending on source and methodology, but at the end of the day, he's averaging the results to create his own forecast.

[Robert Brown]

Does FiveThirtyEight have anything to do with the early bird special?

[oakapple]

FiveThirtyEight is terrible and is ruining the goodwill Silver built up during his stint with the NYT. That piece is a good illustration why.

 

Think of it this way. For years, it has been my hypothesis that the Michelin stars are well correlated with other measures of restaurant excellence. A number of my observations have been consistent with that hypothesis, but I've never tried to test it rigorously. If you wanted to do that, how might you go about it? Silver did exactly that, and I see no flaw in his analysis—as far as it goes. Of course, it's a "drive-by" piece, so I wouldn't take it as proving the hypothesis, but it's another useful set of data points trending in that direction. Understood this way, I see nothing wrong with it whatsoever. If the best @tighe's got is that Yelp rankings aren't a continuous function, I call bullshit on that.

[oakapple]

The star ratings are no different than hype words that try to say everything and end up saying nothing, while the text is a bunch of indisciminate or undiscriminating babble that unlike Trap Advisor or Zagat doesn't even allow for reading between the lines.

What I am trying to understand, is whether you just think that stars are always meaningless, or whether you think these stars are more meaningless than other people's. If it's the former, then your beef is not with Michelin, but rather with the whole concept of using stars—a valid concern, but not a criticism of the guide itself.

 

If it's the latter, you haven't really explained what makes their stars worse than others.

 

The title escapes me, but about ten years ago a former Michelin inspector wrote an exposé about how corrupt the inspection system was and how the inspectors often didn't visit certain restaurants, all of which begs the question as to who these inspectors are, what are their qualifications and past experiences and how they go about their job.

It's true, that exposé was written, but that was a decade ago, on another continent. Obviously, it would be very troubling if it were true today. Nothing I've seen suggests that it is.

 

Also what do ten-year old NYT restaurant reviews have to do with anything?

Because I submit that the Guide doesn't need to be perfect, it only needs to be better than the alternative. The fact that the NYT allows important restaurants to go unreviewed for a decade or more is clearly a demerit of their system, which the Michelin system does not suffer from.

 

The end result of all this is what I have always believed: first restaurants present a bad form, or are unsuited, to be reviewed, and that the only person you should rely on is yourself and the research that you do. You can tell a lot more about a restaurant from studying its website. It's where I go first because I get the best indication available as to whether it fits what I look for in a restaurant and, most important, the aspects of contemporary restaurant dining that I can't abide.

If that type of research works for you, by all means do so. But surely you understand that there are people visiting a foreign city, who don't have the time to read hundreds of websites before deciding where to have dinner. I don't use the Guide in New York. As I live here, and dine out here regularly, it's worth my while to be familiar with the landscape. If I were planning a few days' visit to another city, there's no way I could find the time to do that.

[Orik]

See:

 

http://chowhound.chow.com/topics/985393

[tighe]

 

FiveThirtyEight is terrible and is ruining the goodwill Silver built up during his stint with the NYT. That piece is a good illustration why.

 

Think of it this way. For years, it has been my hypothesis that the Michelin stars are well correlated with other measures of restaurant excellence. A number of my observations have been consistent with that hypothesis, but I've never tried to test it rigorously. If you wanted to do that, how might you go about it? Silver did exactly that, and I see no flaw in his analysis—as far as it goes. Of course, it's a "drive-by" piece, so I wouldn't take it as proving the hypothesis, but it's another useful set of data points trending in that direction. Understood this way, I see nothing wrong with it whatsoever. If the best @tighe's got is that Yelp rankings aren't a continuous function, I call bullshit on that.

 

 

If I only had a dollar for every time an 'IT Professional' has asserted their superior grasp of statistical methods.....

[Anthony Bonner]

LOL