3 ms·

Let me try to make hexane360's point more explicitly: Imagine a hypothetical universe where all Haskell books are equally good, but some are more popular than

by marvy 8y ago

Let me try to make hexane360's point more explicitly: Imagine a hypothetical universe where all Haskell books are equally good, but some are more popular than others (maybe they have well known authors, or catchy titles, or are available for free online, or something). In particular, imagine Haskell Book Number 7 (HB#7 for short) gets 99% "market share". Then if someone says "most people who gave up on Haskell used this book", you should not conclude that it reflects badly on the book. It's just a consequence of the book being popular, so it's the one people tend to try to use, so it's "the one that didn't work for them". If you want to learn whether the book is better or worse than average, you need to split people into two groups: one group used this book, the other group used some other book, and compare which group does better. (Ideally, you should assign people to a group based on a coin flip, but even just looking at what happens "naturally" without your interference is better than nothing.) Back to our non-hypothetical universe: LYAH does not have 99% market share, but it's pretty popular. I heard of it, and I've never looked for a book on Haskell. So to some extent the same logic applies.

NateEag 8y ago
I see what you're saying, and it makes some sense. As you say, though, LYAH is almost certainly not at 99% market share. It's definitely not at the 100% market share HTML has in Web development. The universe is too complex to design a statistically-valid study for everything you want to know (though I definitely understand the urge to do that). At the end of the day, anecdotal evidence may not be great, but it's often better than nothing, and nothing is often the only other practical option.
- marvy 8y ago
  > better than nothing, [which is sometimes the only other option] True. The trick is to sort of mentally keep track of how much better than nothing it is, so you know which pieces of anecdotal evidence you can trust with only 40% confidence, and which you can trust with something more like 90%. But yeah, definitely WAY better than nothing.

by marvy 8y ago

NateEag 8y ago

I see what you're saying, and it makes some sense. As you say, though, LYAH is almost certainly not at 99% market share. It's definitely not at the 100% market share HTML has in Web development. The universe is too complex to design a statistically-valid study for everything you want to know (though I definitely understand the urge to do that). At the end of the day, anecdotal evidence may not be great, but it's often better than nothing, and nothing is often the only other practical option.

marvy 8y ago
> better than nothing, [which is sometimes the only other option] True. The trick is to sort of mentally keep track of how much better than nothing it is, so you know which pieces of anecdotal evidence you can trust with only 40% confidence, and which you can trust with something more like 90%. But yeah, definitely WAY better than nothing.