Statistics
- Thread starter JLM
- Start date
I'll most likely be rebuffed by someone much more knowledgeable on the subject
of Statistics than I am, but Statistics can show pretty much what you want them to
show depending on the parameters set out to display the information.
The whole Climate thing I'm not going to touch, so lets use crime as an example.
Crime rates have been dropping for a long time....everywhere, it seems. Hmmm..
Antidotally, my Father or any of his Friends 30+ years ago never had their cars
broken into or stolen. I'm sure it happened, but not to anyone they knew. Now here
I am and I'm not sure if I know anyone personally that hasn't had a car broken into
or stolen in the last five years (& I'm leaving the never or ever out of this now).
Similar story with B&E's of peoples homes (& many where not locked much of the
time) back 30+ years ago. If you heard of it, it was a rare occurrence. Now I'm hard
pressed to think of anyone I know personally that hasn't had a B&E (or an attempt
was made on their homes) in the last 5-10 years....and so on & so forth. The crime
rate has been dropping for years though according to the statistics. What's going
on?
of Statistics than I am, but Statistics can show pretty much what you want them to
show depending on the parameters set out to display the information.
The whole Climate thing I'm not going to touch, so lets use crime as an example.
Crime rates have been dropping for a long time....everywhere, it seems. Hmmm..
Antidotally, my Father or any of his Friends 30+ years ago never had their cars
broken into or stolen. I'm sure it happened, but not to anyone they knew. Now here
I am and I'm not sure if I know anyone personally that hasn't had a car broken into
or stolen in the last five years (& I'm leaving the never or ever out of this now).
Similar story with B&E's of peoples homes (& many where not locked much of the
time) back 30+ years ago. If you heard of it, it was a rare occurrence. Now I'm hard
pressed to think of anyone I know personally that hasn't had a B&E (or an attempt
was made on their homes) in the last 5-10 years....and so on & so forth. The crime
rate has been dropping for years though according to the statistics. What's going
on?
What's going on Ron is that the human brain is selective, and subjective. Statistics should be objective, and when Stats Canada says that crime rates are dropping, it's fairly certain that this is the case.
If you really want to know why there is a disconnect, think of it this way. Statistics deal with arithmetic means. They take a large sample, or census data, and make inferences about the population based on the mean. You and your friends and families are single realizations, and you will remember an event (car theft, break and enter) more than a non-event (no crime today, did you take notice?)
One part of doing statistics is having proper amounts of information to realistically be able to say something about the population you're interested in. Anecdotes aren't going to cut it.
I know you didn't want to touch the climate thing, but I will say one more thing. A proper amount of information in statistics means you're dealing with sample size. The more values you gather, the more they tend to approach something called the normal distribution (it's the bell shaped curve.) It essentially means that for values far away from the top of the bell, the mean, that the probability of a single measurement is very low. The probability of obtaining a data point close to the mean is very high. It turns out that as you get close to thirty values, many types of data will begin to appear normally distributed. Much statistical analysis relies on the normal distribution to make good reasoned inferences.
Anyways, climate is defined as thirty years, because at this value, the individual data points should appear to be close to normally distributed. This tendency is called the central limit theorem. So climatologists can make meaningful claims about changes in the system when this normality is apparent. This is a highly simplified explanation of course, there's far more to it once you go down the rabbet hole.
If you really want to know why there is a disconnect, think of it this way. Statistics deal with arithmetic means. They take a large sample, or census data, and make inferences about the population based on the mean. You and your friends and families are single realizations, and you will remember an event (car theft, break and enter) more than a non-event (no crime today, did you take notice?)
One part of doing statistics is having proper amounts of information to realistically be able to say something about the population you're interested in. Anecdotes aren't going to cut it.
I know you didn't want to touch the climate thing, but I will say one more thing. A proper amount of information in statistics means you're dealing with sample size. The more values you gather, the more they tend to approach something called the normal distribution (it's the bell shaped curve.) It essentially means that for values far away from the top of the bell, the mean, that the probability of a single measurement is very low. The probability of obtaining a data point close to the mean is very high. It turns out that as you get close to thirty values, many types of data will begin to appear normally distributed. Much statistical analysis relies on the normal distribution to make good reasoned inferences.
Anyways, climate is defined as thirty years, because at this value, the individual data points should appear to be close to normally distributed. This tendency is called the central limit theorem. So climatologists can make meaningful claims about changes in the system when this normality is apparent. This is a highly simplified explanation of course, there's far more to it once you go down the rabbet hole.
Yeah....I wasn't going to touch on the climate thing, 'cuz there's a couple of threads
here topping the 2000 post mark on it, & they're still going on...so that's why I went
with the crime thing.
Yeah, the Human mind is a weird thing. I'll agree with you there. ;-)
Can statistics be manipulated based on a predetermined concept (I'm NOT talk'n
about the climate) by setting up the parameters governing the display of the stats to
reflect what someone wants them to reflect?
Not talking about the Climate thing or the Crime thing, look at the whacky conflicting
statistics thrown around regarding the whole Gun Control thing....What'cha figure?
here topping the 2000 post mark on it, & they're still going on...so that's why I went
with the crime thing.
Yeah, the Human mind is a weird thing. I'll agree with you there. ;-)
Can statistics be manipulated based on a predetermined concept (I'm NOT talk'n
about the climate) by setting up the parameters governing the display of the stats to
reflect what someone wants them to reflect?
Not talking about the Climate thing or the Crime thing, look at the whacky conflicting
statistics thrown around regarding the whole Gun Control thing....What'cha figure?
You can use me as someone you know. Had my car broken into in the early 80s and have not had my house broken into ever.
We all know stats are as reliable as they need them to sound. I've said it over and over, stats are changed all the time to suit the needs of the person needing to show them. My husband was listening to a report on how crime has dropped across the country and he's one of those people who seems to think that if he speaks loud enough to the TV, someone will hear him so - he's yelling across the room - That's wrong, that's wrong. All they can do stats on are reported crime and many crimes go un-reported! I certainly heard him.
Sure they can. Just like most anything else, if someone wants to abuse statistics they can. Statistics is just a tool. A very powerful tool if it's used correctly. People in the know will probably speak up if someone is using statistical methods improperly, or being fraudulent.
From my field, if we want to test a new vaccine, we don't use more fish in the trial than we need to. For one, it's an animal welfare issue, and second, using more fish would tie up resources that can be used in other trials (tank space, fish, heated or chilled water, etc.) We use statistics to determine how many fish we need for statistical significance.
Telling the whole story?
Zero reliability.
Statistics can show an accurate view towards one particular aspect of a much bigger picture, but almost always, statistics do not account for every needed factor, or bring about an assumption of something unfounded based on the information provided that with an additional factor applied, could bring about a totally different result.
It might show information leans towards one particular thing, but does it explain how it got there, why people decided on something, why something is the way it is in its entirety, the background or the reasons someone or something came to such a conclusion?
Never in its entirety.
Statistics are good for giving a limited perspective on a subject and perhaps point people in the right direction in regards to gaining further and more detailed information on said subject, but one should never rely on a statistic as being 100% true or that it tells a story/situation in its entirety.
Besides, 87.62% of all statistics are made up on the spot ;-)
Ok...can you name anything which we humans have come up with to describe things that can account for every factor? Sometimes we're not interested in every factor.
No, statistics is just a tool. You don't build an entire house with a hammer, but it does come in handy for some parts of the build. Statistics is no different.
Statistics uses probability. There are some results which are so solid, that the notion that they came about by chance are so improbable as to be negligible.
If you try to make more out of the stats then they can give, then you're using them improperly. Despite what you all might think of stats, the truth of the matter is that the progress we've made in science would have been a fraction of what we have discovered and learned if we did not use statistics when analyzing scientific results.
One poster gave a good example of the fallibility of statistics, in that for crime statistics, not all the information is entered because not all of it is even reported.
Statistics are only as good as the quality of the data. Caveat.
But now here's a pickle for you. Do you think that this non-reporting of crime is a new phenomenon, or something that has always been there? If so, showing a decreasing trend in crime rates is very likely to be a robust result. If not, then the statistics which show that this is a new phenomenon lead to investigations, which can find ways of addressing the problem.
... and the description of what the stats are supposed to show.
Interesting thing in BC lately: someone did a study on the public's view of the RCMP and found about only 35% of the public had confidence in the RCMP here. I can imagine that this stat would cause a few people to just suffer minor crimes and keep clam about them rather than reporting them. That could cause a lowering of crime stats and reporting. Ironic, huh?
Are five-day weather forecasts 100% accurate? There's statistics in action.
Around twenty-five years ago, I was dumped by Wawanesa Insurance because, according to their statistics, my luck was soon going to run out. It hasn't happened yet - and Wawanesa cheated themselves out of a quarter century's worth of free money.
Statistics? If we lived in a world without change, they'd have a better chance.
Around twenty-five years ago, I was dumped by Wawanesa Insurance because, according to their statistics, my luck was soon going to run out. It hasn't happened yet - and Wawanesa cheated themselves out of a quarter century's worth of free money.
Statistics? If we lived in a world without change, they'd have a better chance.
What was the % before say... the tasering incidents? That's a crucial bit of information to have before anyone makes any kind of inference.