Abuse of Statistics

As usual, the Marketing wonks rely on the people's lack of understanding. There are lies, damn lies, statistics, and then, the worst of all, MARKETING! The biggest marketing lie of all is the one when you are told that eyeballs = sales. The more people see your company name, product, logo, whatever, the more you will sell. Utter rubbish. Think back to your favourite TV ad. Come on, everyone has one….mine was the Finger of fudge ad with the kid playing conkers with his mates, closely followed by the 'Wazzuuupp' Lager add. Did the adds make me go out and by any of these products? No, in fact I can't even remember what the lager was! Adds are entertainment. End of story.

Baaad Kitty

Abuse of statistics in medical news is more common than you might think. The only way to have a clear perspective on the headlines and media bytes is to look at the study itself, and that is not always so readily available.

Here's a good example, where the stats are coherently reviewed online for the benefit of doctors at http://www.medpagetoday.com/Neurology/MultipleSclerosis/8341 The study was published in a peer reviewed journal, so that in itself is no assurance of legitimate statistical methods.

Many of the media bytes stated that marijuana caused cognitive impairment in MS, on the basis of this study. The review shows that the headline was based on an assessment of ten of 140 patients who self-reported mj use ("at least once in the previous month". They were never tested for cognitive function when they were actually using mj. (MS itself causes cognitive impairment). The ten were 'matched' with four self-reported non-users each, in order to bump the numbers up to 40 for statistical purposes, avoiding the crappy confidence intervals of the student variable.

Oh, and, note the conflict of interest of the researchers, disclosed at the bottom.

Since the number of the 140 MS patients who admitted to ever smoking mj is well below the statistics reported for the general population (http://www.statcan.gc.ca/daily-quotidien/040721/dq040721a-eng.htm), the study may as well have concluded that the ten who self-reported mj use to a psychiatrist, did so because they were more cognitively impaired by MS than the patients who (statistically speaking) may have decided to lie about it in this context... and who may even have been included in the sample of 'matched' 'non-users'...

Or does it show statistically that people who have never tried mj are more likely to get MS??? heavens...we're talking at least 10% greater risk according to what is shown in this study... why aren't they screaming this warning to the public.....

Talking of small samole size...I once was in a meeting and reported a single instance of something...a marketing man jumped on it and asked if this was a trend!

Del

There was a Parliamentary quote a few years back, somewhere along the lines of, "The rate of increase of inflation is increasing" _{(Stan is going to sue your **** !)}

And they say that there was a statistician who drowned in a lake with an average depth of 2 inches.

Brilliant !!

62% of all statistics are made up on the spot. Of these 48% can somehow be made to seem accurate. With enough fudging and rearranging of statistics, 100% can be made believable to the public.

This is a perfect example, as I have no idea what I just said!

What about weather forecasts? What exactly is a 50% chance of rain? I know what a 50% chance of "heads" or "tails" on a coin flip is but rain.......

Another obvious example is mostly any sports statistic. Pure rubbish.

LOL - you've been watching the Pessimist Channel !

In the UK, cold weather is reported in Centigrade, and hot weather in Fahrenheit. It gives us more 'wow' factor on our favourite topic.

Are, I know this one! 50% chance of anything means; 'I don't know'. Easy.

50% chance of rain can have very clear meaning:

Non-statistical: 50% of the area covered by this forecast will experience rain.

Statistical: if you were to hear this forecast an infinite number of times, the average occurrence of rain over the area covered by the forecast would be 50%. (So, a mix of uncertainty and local variability - if you ask pay for a forecast for a particular location they can often do a lot better than that)

If the forecast comes from the MET office (or one of its commercial offshoots), I would normally assume that the originator had reasonable understanding* of statistics and was using the terms properly - and also relying on people learning from their experience of what happens following similar forecasts (if only my memory were good enough...).

Of course you could satisfy the second criterion by always forecasting that the probability of rain was equal to the average for the time of year; and even that could be of some use. But actual weather forecasts are appreciably better than that - albeit living where I do I can reduce any probability given for rain by about 40% (of the given value)

*= what you do with an umbrella during a rainstorm

50% chance of rain can have very clear meaning

Not for me.......It could be planar sheets of water falling over the whole area, with nicely timed intervals between them. It might seem inapt for me to say it, but I prefer to think of cold air sinks, rather then the popular 'hot air rises' stuff.

I'm too lazy to retype this oldie, but you'll know it I'm sure.....

An astronomer, a physicist and a mathematician were holidaying in Scotland.
Glancing from a train window, they observed a black sheep in the middle of a
field."How interesting," observed the astronomer, "all Scottish sheep are black!"To which the physicist responded, "No, no! Some Scottish sheep are black!"The mathematician gazed heavenward in supplication, and then intoned,
"In Scotland there exists at least one field, containing at least one sheep,
at least one side of which is black."

No, that's rain falling 50% of the time - and if you make your sampling area small enough, that's not true even in a heavy rainstrom. The forecast usually covers the day ahead - so a 50% chance of rain is meant to be the chance of it falling at some time (which is why forecasters usually qualify their statements). Of course, it might just be that showers in MK last longer than elsewhere...

Now, which of those three do you think applies?

I didn't see Sophysicist in there ?

"What exactly is a 50% chance of rain?"

It is the 100% probability that only 1 out of every 2 raindrops will fall. The rest merely hover.

"only?" If we count droplets that are potential rain, the probability is much smaller than that. If we only count those that fall...

To be temporarily and apologetically serious:

It is very difficult to find estimates of spatial and temporal probability density functions of rain volumes. Especially so for small (1-5 meter, 1-10 second) scales.

Actually, there's a 50% chance of rain tomorrow 100% of the time - either it will or it won't!

That is, there's a 100% chance of a chance (of rain) which itself can assume values between 0% through 100%? Sort of like Probability's version of a second derivative?

You can find some slightly topical stuff here, Del.

eg;

The passive smoking report shows a link between passive smoking and sensitivity to ragweed pollen. Syringing your ears makes you, statistically, more sensitive to sound. If you send someone with freshly syringed ears into a noisy environment he will suffer more hearing loss than someone whose ears have the natural protection of earwax.

This shows that statistically syringing ears is damaging.

If I had asked my volunteers to cross a road blindfold, I would have shown that syringing ears is a good thing, as hearing oncoming traffic is an advantage, but would have proved conclusively that listening to, and following the advice of, men in white coats is damaging because walking across a busy road blindfold is exceedingly stupid.

And I thought 100% of all cat reference posts were Del inspired. What is Kris doing? Does this prove Kris is Del, or that Del is Kris, or that I have thought a statistic shows something it doesn't.

Simon

.....It was a slight mishap in developing the KrisDel™ Time Machine. Think "The Fly", but an awful lot worse.. It can be tricky to tell us apart though - Del's got good grammar/spelling, and people say they can't recommend mine highly enough.

No, no, that was a misquote from the intrinsic safety forum: "I can't recommend mining highly enough"

How low can you get ! I'm constantly trying to find out

Well, it servers them right for smoking ragweed in the first place! Stupid rat-bastards!!!

Why don't we ask a REAL scientist...

There's a thought - I wonder how many people who like the occasional Moroccan Woodbine don't consider themselves smokers ?

That Scientist shouldn't be allowed in - he's not carrying his certificates. Oddly enough, he can get in if he talks utter balls. What a conundrum . 5 times out of 4, people who introduce themselves with suffixes are spouting crap.

"...5 times out of 4..."

I'd have guessed it's at least twice that...

Now I'm all stuck on 10/8 or 6/4 !

10 times out of 4 it's gibberish by blatherskites! They refuse to eschew obfuscation, and it's frustrating.

"Contrafribularites ? .........Oh, I'm sorry, sir. I'm anispeptic, frasmotic, even compunctuous to have caused you such pericombobulation "

I don't know if this would qualify as an abuse of statistics necessarily, but last week it was reported that a Statistics professor from one of the local universities drowned in a lake having an average depth of one centimeter.

How many times have we seen the efficacy of anti-cancer or now anti-aids or other mediceines - the xyz medicine released by abc company reduces the incidence of cancer by 300%

(small foot note the medicine and the placebo was used on controlled groups the incidence of testicular cancer with placebo was found to be 3 / million and on the xyz was 1/million hence...)

Or few months back i have seen an article in British medical journal - rather it was reproduced in my newspaper -chances of cancer more with cell phone than with smoking (thankfully i belong to the rarest of species that smoke)

best statistics i have seen ever are ofcourse from the most reliable agencies - the governments and the oppositions.

The working is possibly the least of the problems - the whole concept is flawed.

We can assume that the archer in question was familiar with his 'best' kit. But we don't whether he had the opportunity equally to familiarise himself with the 'beginner' kit. Let us be generous and assume that he did, and that the size of the experiment was adequate to show that these would be the expected scores (and let's not worry about placebo effects for now?)
What conclusions could we reasonably draw?
Possibly that even with a "perfect" aim, the variability in the dynamics of the beginner's kit will limit your score to 92/100 for the particular conditions under which the archer was firing.
And... does that mean the beginner would be better off with 'best' kit? Probably not - assuming this was a longbow, the higher tension of the drawstring of best kit would probably be more than the beginner could cope with.

I can truthfully say that "I gave this more than 500%, innit". (100% being the effort that it would be productive to devote to the activity). (your 'semi-literate sportsmen' may be sharper than you realise)

Tiger Woods could play golf with Del the Cat's clubs, and probably shoot a pretty good game. Del could play using Tiger's clubs...and...still...be...a...duffer...

I was once having a discussion with my boss, the Technical Director, during review time. I told him, "Do you realize almost half of your staff is below average?" He said,"Yes, I know, but I'm trying to improve that."

When you consider that almost half of the voters are below average in intelligence, it almost explains the mess we're in.

And everybody is so proud that their kids are "above average".

Almost half of me is below average

That's like Garrison Keillor's fictional town of Lake Woebegone, Minnesota, "where all of the children are above average".

It is widley know amongst staticians that 72.45% of all statistical data is made up on the spot.

You are underestimating them and talking about 9.273% of them.

Just beware - and be ready for a legal notice - your e-mail address is anyway there with CR4

Food for thought : A legal notice can be served on E-mail ?