Wow, the Voyage is an awesome ride. No doubt deserves to be in top 10 at least. My closest wood coaster is Gwazi and that sadly wouldn't make my top 150!
All it gave me was a bunch of text. I think the site is messed up. There's no poll to even take. The day I was at SFNJ El Toro wasn't open and I haven't ridden many of the newer top wood coasters to be a fair judge, but was seeing which coasters were listed on the poll. I still miss Son of Beast.
CoasterGoddess said:
All it gave me was a bunch of text. I think the site is messed up. There's no poll to even take.
You have to actually read that text.
Oh.
Emm.
Gee.
Here's the results: http://www.ushsho.com/woodrollercoasterpollresults2013.htm
I can't be bothered with all that. Where are the pictures? Give me a big ass number with a rollie coaster next to it.
I would write about the statistical insignificance of ranking rides with 17 opinions versus hundreds, but that's like explaining science to Michelle Bachmann.
Jeff - Editor - CoasterBuzz.com - My Blog
But it's plenty significant for the application.
http://coasterbuzz.com/Forums/Topic/mitchs-2012-wood-coaster-poll/2#874967
And as I say every year, you don't get to make up the rules for hundreds of years of accepted standards for statistics. Winky and such.
Jeff - Editor - CoasterBuzz.com - My Blog
The average number of wooden coasters ridden was 31.4. That surprised me a little, I would have put that number closer to 50. Not sure what that says about the poll....or about me.
And as I say every year creating an incomplete list because you're hung up on a margin of error for something as silly as a roller coaster list is actually creating a less accurate list.
No winky. Getting hung up on the details is causing you to miss the forest for the trees.
To me, a complete list using insufficient data trumps an incomplete list using sufficient data. It has nothing to do with accepted standards and everything to do with creating a complete list.
"That coaster can't be the best because only 17 people rode it" is every bit as wrong an answer as you feel using the insignificant data is.
Both approaches are technically incorrect as one ignores subjects for lack of quantity of data and the other ignores lack of data for subject inclusion.
They're both equally wrong. It's like Andy said, it's up to the individual to decide which way to be wrong.
A list ranking all the roller coasters that doesn't actually include all the roller coasters is every bit as broken as you claim one that uses statistically insignificant data...probably more so.
Would you change your mind if you read a book on statistics? I didn't make this stuff up, it's from centuries of people much smarter than me.
Jeff - Editor - CoasterBuzz.com - My Blog
I'm not disagreeing with the scientific method (for the umpteenth time).
I'm disagreeing with applying it so rigidly to a list of roller coasters.
To me, the goal is to create a complete list. Using statistically incomplete data does that better than ignoring entries altogether.
You're hung up on the methodology and it's making your (hypothetical) list less accurate by doing so. Assuming even remotely reasonable methodology, a list that includes ALL the roller coasters will be better than one that doesn't...every time.
The statistically insignificant entires on the list might not be scientifically correct, but with a little bit of common sense and interpretation, they list just might be dead on (or at least close).
The second you ignore a single coaster and exclude it from your list, your list is wrong. It cannot be an accurate ranking of all roller coasters. It's an accurate ranking of some roller coasters...but that's even more useless than a list ranking roller coasters already is.
And it doesn't need a book or historical precedence to make that point. If I ask you to rank 10 things and you give me a list back that includes only 8 of those things - you failed. There's no denying that. Even if you didn't know the 'right' way to rank those 10 items, an educated guess would be a better option than just skipping some of the items on the list.
Daft Punk won album and record of the year, El Toro got tops on Mitch's poll, and Justin Beiber keeps getting arrested. All in all a great week.
Lord Gonchar said:
To me, the goal is to create a complete list... Using statistically incomplete data does that better than ignoring entries altogether. The statistically insignificant entires on the list might not be scientifically correct, but with a little bit of common sense and interpretation, they list just might be dead on (or at least close).
But statistically speaking, they're not even close. It's like saying because I married a redhead, all women have red hair. If the goal is to have a complete list, you need a bigger sample. We've talked ad nauseam about issues of selection bias and sample size, and how they produce less accurate results. For anything in statistics. You can introduce all the common sense and interpretation you want, but it reminds me of Jenny McCarthy and her nutty "vaccinations cause autism" crusade.
I'm not really hung up on anything, I've just spent a lot of time studying polling techniques (not by choice, mind you... it was in a college journalism class). Not surprisingly, most TV news polls wouldn't pass basic scrutiny either.
Jeff - Editor - CoasterBuzz.com - My Blog
Jeff said:
But statistically speaking, they're not even close.
Yeah, the statistically 'correct' one ends up incomplete and essentially useless for the intended purposes.
It's like saying because I married a redhead, all women have red hair.
No, not at all. It's like saying you have 12 good friends. 6 married blondes, 4 married brunettes and 2 married redheads. Therefore blondes must be most popular, brunettes second most, and redheads third.
It's not even close to statistically significant, but it's an entirely reasonable guess given the data (and it might just be right). Your list would just exclude redheads because there wasn't enough data. And it would be immediately incomplete (and incorrect) because of it.
If the goal is to have a complete list, you need a bigger sample.
To have a statistically significant complete list you need a bigger sample.
It still doesn't mean the statistically insignificant list is "incorrect", per se.
And I still argue that given the fact that poll uses comparisons, the samples aren't just who has ridden any given coaster, but rather who has ridden any given pair of coasters.
And if you're talking a couple of pretty obscure coasters, you essentially have Richard Bannister and a handful of other hardore travelers qualified to make that comparison. If only a dozen people have ridden a given pair of coasters, you've effectively sampled at a rate of 100%. It's literally impossible to obtain a bigger sample. You have as much data as to how those coasters compare as is humanly possible.
Is it insignificant? Possibly. Is it the best sampling that anyone can do. Yup.
You can't sample more than exists.
You can introduce all the common sense and interpretation you want, but it reminds me of Jenny McCarthy and her nutty "vaccinations cause autism" crusade.
Jenny McCarthy is a wacky old broad with no understanding of how vaccinations work. Not at all qualified to give medical advice.
My guess is that Mitch Hawker is a guy with a decent understanding of roller coaster opinion at this point. As qualified as anyone I can think of to intepret lean data on thrill ride rankings.
That's the difference. Don't discount the 'art' of presenting science to the masses.
If you say so. Statistical significance is not really a concept open to interpretation. You either have it or you don't.
Jeff - Editor - CoasterBuzz.com - My Blog
You must be logged in to post