Why Librans' cars always break down

[Sticks]

:eh:

From the Daily Mail online (http://www.dailymail.co.uk/pages/live/articles/news/news.html?in_article_id=524320&in_page_id=1770)

If you're a Libra driving a blue car, better make sure you have plenty of breakdown insurance.

For statistically you're the most likely to end up stuck at the side of the road waiting for help.

Research among the RAC's one million members suggests that Libran drivers - those born between September 24 and October 23 - fall prey most often to mechanical failure.

Is this another one from the "Meaningless Statistical correlation" department?

[Laguna]

Good that I am virgo and have a black car...
And a mobility insurance, and still two years warranty left...

[NEOWatcher]

While offering few scientific explanations for its findings, the RAC survey does come up with some interesting statistics - particularly for the superstitious.

Few? How about mentioning at least one?


a whopping 42 per cent have read their horoscope in the past year compared to only one seventh (13 per cent) who have read the Highway Code.

Oh, like that's fair. The horiscope changes daily, while the code changes...:wall:
(ok, could be daily, but I'm sure the published changes are much rarer)

[Ivan Viehoff]

I'm a Libra with a blue car, but unlike any journalists at the Daily Mail I have some basic training in statistics. The RAC say:

"RAC internal data shows that Librans are 1.46% more likely to break down than Pisceans [the least likely to break down]. 1.46% of RAC's 86,848 Libran members equates to 1,268 people. The data is based on 1,072,209 RAC members and there are 34,775,587 UK motorists. Therefore there are 41,121 more Libran breakdowns than for Pisceans."

(Let's ignore the fact that the final sentence is just nonsense.)

I'm trying to work out whether this difference is statistically significant, which depends upon the size of the sample. They don't say what size the sample is.

Lets try and guess one. The AA suggests that an average weekday sees about 26,000 breakdown call-outs, so I reckon about 9 million a year. So that is about 1 in 4 motorists makes a call out in any given year. So about 250,000RAC members make a call-out in a given year. So I'm going to take their sample size as 250,000, which I'm guessing is the number of callouts to their members in a year. (Interestingly, the RAC say that they supply about 2,500,000 call-outs in a year, so presumably they are also substantially giving service to non-members; it indicates a market share of about 35% which sounds about right, but they are unlikely to know the birthday of non-members).

The chance of being of a given star sign is 1/12. The standard deviation of a binomial distribution of probability 1/12 and 250,000 tries (variance is np(1-p))is 138 which is 0.66% of the mean (np). So Libra and Pisces are 2.2 standard deviations apart. Which if anything sounds on the low side, for the difference between the highest and the lowest of 12 numbers drawn from a normal distribution.

Although I don't know their sample size, what I read doesn't sound like anything different from purely random effects of no statistical significance.

[Sticks]

In other words the answer to my question in the OP is Yes?

[HenrikOlsen]

Since they gave probabilities without giving standard deviations and with no indication of how they came to looks at exactly that subdivision, the answer can be seen to be yes without doing any calculations at all.

This is one of the classic faulty uses of statistics.
Take a large sample space,
Loop:
divide sample space into arbitrary subsets
Check all subsets and record difference between expected average size and actual size
If all differences are too small repeat loop with new subsets
Publish found large difference as exciting news.

[Tobin Dax]

I finally decided to open this post and see what trouble British librarians were having with their cars. :doh:

[Mister Earl]

I finally decided to open this post and see what trouble British librarians were having with their cars. :doh:

Same. I even had a witty retort warmed up and ready to go.

"The reason Librarian's cars break down so often is to force them to read the automotive books they've been neglecting to read for so long."

Self fail.

[danscope]

Well, I'm a libra and I also have a very good mechanical background. It has always been my mechanical philosophy to go over my vehicles with a view to
ensure I don't have problems. When I buy a vehicle , I remove all the fuel , and change the filters....ALL the filters. I get a new and very heavy duty battery.
I keep my tank topped up so it doesn't acqure a lot of water(moisture condenses and NEVER evaporates once it is on the bottom.
When you are thorough about your vehicles, you can drive with confidence.
The generalization about astrological signs and break downs is poppycock....
rubbish from someone sitting in front of a word proccessor.
Take care of your car and.....Press On !!!
Best regards, Dan

[HenrikOlsen]

I see I forgot two steps in my recipe of fame, here is the revised one.
Take a large sample space,
Loop:
divide sample space into arbitrary subsets
Check all subsets and record difference between expected average size and actual size
If all differences are too small repeat loop with new subsets
Ask a statistician what the probability is for this to occur randomly
(Important note! Don't tell him you picked it out because it was different)
Make a hypothesis explaining the difference, using as many new-age buzzwords as possible.
Publish hypothesis as exciting new theory, making sure to quote the statistician as evidence for theory.

[trinitree88]

SNIPPET... I keep my tank topped up so it doesn't acqure a lot of water(moisture condenses and NEVER evaporates once it is on the bottom.

danscope. The function of drygas is to supply the tank with a molecule that has a hydrophobic/ oleophilic end, and an oleophobic/hydrophilic end. A short chain alcohol does the trick...it attracts the water molecule with the alcohol group, and is soluble in gasoline with the alkyl group, so it carries the water to the combustion chamber to be vaporized and exhausted. There's no fuel value in doing so, but it prevents emulsified waters from traveling through cold fuel lines in winter and accumulating as ice in the fuel line where it is exposed to severe wind chills. So although the water can spot collect, a regular dose of drygas between Halloween and Easter, say 1 can per tankful ought to easily alleviate the problem in cold climates....doesn't affect your octane rating a lot, and if your gas is 10% ethanol you don't need it at all.
pete

[danscope]

I agree. I still drain the tank, though.
Happy driving and happy-go-lucky are indeed, two different things.

Dry gas never hurts.

Best regards, Dan

[Ronald Brak]

Is this another one from the "Meaningless Statistical correlation" department?

Yes. Some group had to be the one most likely to break down. It turned out to be Librans in blue cars. It's about as earth shattering a finding as the discovery that half of nuclear reactors have below average safety records.

[mfumbesi]

Well Ron you are making me nervous now.(Its in jest I hope)

[Ronald Brak]

Well Ron you are making me nervous now.(Its in jest I hope)

On more than one occaision I've told people there is no way to absolutely prove anything beyond all doubt, but on this point I am completely certain. One half of all nuclear power stations have below average safety records. However, it's not quite as bad as you might think. In fact, it's pretty much unavoidable.

[Delvo]

One half of all nuclear power stations have below average safety records. However, it's not quite as bad as you might think. In fact, it's pretty much unavoidable.For anyone who didn't catch what he's doing here, the idea is that half of any group of anything is below the average in any aspect of them you measure, and half of the group is above, because that's what an average is: the point at which half are above and half are below. You could also say that half of the human population is fatter or less intelligent than average.

But that's only true if the distribution curve is symmetrical. "Average" is not defined as the point at which equal numbers occur above and below; that's the median. With a symmetrical distribution curve, the average and the median will be in the same place, but if the population is generally bunched up toward one side of the curve or the other, then they can separate. For example, among the numbers 1, 1, 2, 2, 3, 4, 6, 9, and 10, the average is 4.2222..., but the median is 3, and there are twice as many examples below the average as above it.

[HenrikOlsen]

BZZZZZZT Wrong! Common mistake about averages.
The value you're describing is the median.

Simple example:
1, 20, 21, 22, 23
Average is (1+20+21+22+23)/5=17.4, one value is below average, four is above.

[Ronald Brak]

"Average" is not defined as the point at which equal numbers occur above and below; that's the median.

BZZZZZZT Wrong! Common mistake about averages.

Yes, that's true. I shouldn't have said what I said. Thanks for pointing it out. Now I'm trying to think of a way to reword the statement so it's true but still has the capacity to freak people out. Hmm... Difficult.

[Torsten]

I've thrown a similar line at people before too, but I've usually prefaced it with a statement that many things we measure have a bell-shaped distribution, and therefore . . .

People can relate to "bell-shaped" easier than "normal distribution". But a lot of people still don't get it.

[NEOWatcher]

Yes, that's true. I shouldn't have said what I said. Thanks for pointing it out. Now I'm trying to think of a way to reword the statement so it's true but still has the capacity to freak people out. Hmm... Difficult.
It did work for me, but I also look at the context.
When it comes to nuke safety, there is the issue that they should all be pretty much the same with the peak near the top of the standards. This means there's lots of room to go down, and very little room to go up. If there is one that is bad enough to throw the average off of the mean, then we got a much bigger problem than to worry about what the average is.

[Ivan Viehoff]

In the case of nuclear power stations' safety record, it depends rather how you measure it. For example, if the measure we are using is "how many times has it blown up", then I think only one US nuclear power station has a below average safety record.

A lot of distributions we come across, we can reasonably expect that the number that is below average is probably not going to be very different from 50%. I do expect that close to 50% of school pupils will have a below average academic record.

Other distributions we do expect to be skewed. I expect somewhat fewer than 50% of people to be more than average weight, because it is much more possible to be double or three times average weight, but it isn't possible to have negative weight. I expect most days to have below average rainfall, because infrequent heavy rainfall days contribute substantially to the total rainfall. Indeed, in many places most days it hardly rains at all. Rather more than 50% of people have below average income, because a few super-rich people increase the average income substantially. Etc.

[Delvo]

Other distributions we do expect to be skewed. I expect somewhat fewer than 50% of people to be more than average weight, because it is much more possible to be double or three times average weight, but it isn't possible to have negative weight. I expect most days to have below average rainfall, because infrequent heavy rainfall days contribute substantially to the total rainfall. Indeed, in many places most days it hardly rains at all. Rather more than 50% of people have below average income, because a few super-rich people increase the average income substantially. Etc.My favorite one of those, because of the spin that's put on it about how we're all supposedly going to react, is penis size. Every "study" yields a smaller answer for the "average" than the last, and the report is always accompanied by comments about how this news is bound to relieve and cheer up a lot men, because it would mean that their own are larger in comparison to average than they thought. But the only way the news of the lower average would have that effect is if it's not true! A low average crowds most of the population between the average and the minimum, so most of the population would still be lower than average (a phenomenon which gets more drastic, farther from a 50/50 split, the closer the average is to the minimum), so the new reported "wow it's so small" average would STILL be higher than most listeners' own measurements, so most guys would be hearing news reports talking about the amazingly small size of an average that's STILL higher than their own. For most guys to think the news is favorable to them, they'd have to hear an average number that's at least as low as their own or lower, but that requires a distribution curve which is, in reality (as opposed to the reports of the "studies") balanced or skewed to the right, with half or more of the population crowded between the average and the maximum. So they're describing the results in a way that means it must be left-skewed but predicting a reaction that depends on reality being right-skewed (or at least less left-skewed than what they're reporting); they're reporting a low average but predicting a reaction that would only happen if the real average is actually higher than what they're reporting.:D:rolleyes::liar::shhh::D