Fluffy thread ... odd take on statistics

iVillage Member
Registered: 03-26-2003
Fluffy thread ... odd take on statistics
27
Tue, 05-13-2003 - 11:32am
Any of you familiar with Todd Snyder? (virgo, I'm sure you are!)

He's a singer/songwriter/comedian. This morning I heard one of his new songs ... and it was a farce on statistics .. and the opening line is something like "60% of statistics are made up on the spot and 86.4% of people believe them whether they are true or not"

I just thought that was an appropos line, given our recent debates, and the focus on studies/statistics on this board.

The rest of the song is hilarious. Goes on to spout 65.4% of this, 98% of that, 23.4% this ... and ends with something like 70% of statisticians hate their jobs ... lol

Hollie

Pages

iVillage Member
Registered: 03-25-2003
Tue, 05-13-2003 - 11:38am
It's silly isn't it! What's crazy is I'm on my way to becoming an actuary, whose primary job is statistics and probability, while I don't put much stock in statistics! (On an individual basis.. although they are VERY helpful in an insurance situation! haha)
iVillage Member
Registered: 11-20-2001
Tue, 05-13-2003 - 1:23pm
I'll be looking for the CD, lol, thanx Holly. n/t

 

Linda - wife, mother, grandmum                     &nb

Avatar for mjdphd
iVillage Member
Registered: 03-26-2003
Tue, 05-13-2003 - 1:33pm
I have taken some statistics courses in my time. They are used to spot trends. But the statistics are only as good as the statistical plan and the interpretation. Ideally, you should have a large random sampling to make the statistics any good. Also, you should have some idea of the margin of error involved in each point. Before anyone starts putting stock in the results, you should research how the results were obtained. People make up statistics all the time to emphasize a point.

Also, something related, is the use of the work average. This is usually the mathetmatical average of a whole group. That often means that very few points actually fall on the average. Anyone in the group can be extremely far from the average as a matter of fact. If you look at the standard deviation, then you will know how large a deviation to expect in a normal situation.

Just my 2 cents. The song sounds funny, BTW.

iVillage Member
Registered: 03-26-2003
Tue, 05-13-2003 - 1:47pm
And, stochaticism is a big problem too.

For those of you unfamiliar with stats (I wish I was! lol), stochaticism is when you have a preconcieved notion of what the results of your research will be, and therefore, your research is structured (purposefully or not) to achieve that preconceived result. Somtimes its even more obvious ... Instead of asking "How does a mother's education level affect her work status?" one might ask "Is a mother without a college degree more likely to SAH?" The 2nd version of that is stochaticism.

Anyhoo ...

Hollie

Avatar for mjdphd
iVillage Member
Registered: 03-26-2003
Tue, 05-13-2003 - 1:57pm
Absolutely. You can structure your test to comeout the way you want it to. That is another reason why you have to research the study before believing the results.

BTW, thanks for a new word. I hadn't heard that one yet.

iVillage Member
Registered: 03-27-2003
Tue, 05-13-2003 - 3:30pm
I liek the average height stats. Just to show that some stats mean *absolutely* nothing irl, I'll give you the average height of my son and his closest friends in 6th grade.

5'4". OK, so some were under 5', and one was 6'. NO ONE was 5'4". So, don't design clothes based on that stat. Don't build doorways based on it. Don't design ergonomically correct furniture on it. Don't design sports equipment based on it. Above all, don't determine your "normalness" by it.

The stat is completely useless IRL. It's an interesting number, but what are we supposed to do with it? Decide that our kid is either short or tall? Since no one is 5'4", no one is average. Every kid is either short or tall. lol

I *have* to get that song.

Avatar for mjdphd
iVillage Member
Registered: 03-26-2003
Tue, 05-13-2003 - 4:47pm
You see, I don't take it as useless data. You could use it to track the height of children over the course of a century to see how people do with varied nutrition. Or compare to children in other countries to help determine nutritional needs.

People do design clothes and other equipment around the average height. Take cars, for example. I am 5'5 1/2" tall. I fit perfectly into cars. The seat, headrest, and pedals are all positioned perfectly for someone my size. My dh is 6'3". He has a hard time fitting into some cars (especially Japanese cars). But since cars can't be made custom built to each person, they have to be able to fit MOST people. And, they do. My dh would argue against that, and so would my 5' tall mother, but for a large percentage of the population, it is fine. Most people probably fit into the 5'3 to 6'0 range that most cars are designed for (I am making up numbers, but you get the point). Yes, people fall outside that range.

Clothing manufacturers have to use some standard to make clothes so that we can have off the rack clothes. It's either that or everyone has to have tailor made clothes. They use a standard that works well for a great percentage of the population.

No, you cannot base normalness on it, but you can design clothes with it. It is not a completly useless stat. If you don't use it, how do you determine how to build doorways, sports equipment, etc? The more sizes that you have to make, the more expensive it is going to be.

iVillage Member
Registered: 03-26-2003
Tue, 05-13-2003 - 5:06pm
You are mixing up proper scientific method and proper design of experiment with proper statistical analysis and interpretation.

As a starting point, this:

"How does a mother's education level affect her work status?"

Is not much better than this:

"Is a mother without a college degree more likely to SAH?" The 2nd version of that is stochaticism.

You want

"Does a mother's education level affect her work status?".

In fact, throw it at someone who really does this stuff and they can probably get it even less biases. Maybe

"Does a persons education level affect their work status".

Because if you can collect that data (with proper design of experient of course) you can then get much more of what you need with the followup statisical work up...because quite honestly, unless you do, you are undertaking the project with a preconcieved bias that a mother's education level affects her work status differently than does the education level of any other kind of person.

Regardless of which of questions you ask, you should get the same information out wrt the issue of mothers - if design of experiment and subsequent statistical data analysis are performed correctly. The data collected in the last example though will answer more questions, and will go a long way to elucidate the relevance of the mother information.

Its not usually the statistics which are wrong, and quite often the scientists and statisticians involved are quite level headed about it. The problem is with the general public - who are largely uneducated as to what they mean and how to intepret and apply them.

iVillage Member
Registered: 03-26-2003
Tue, 05-13-2003 - 5:24pm
Sigh.

How close are you to actually being an actuary?

<>

Sigh again. Look you *can't put much stock in statisitcs on an individual basis*. It makes no sense. One data point doesn't mean a thing - its the whole POINT of using statistics. To be able to predict in general, when the exact future behaviour of one element is not known and cannt be know with certainty...unless you are wondering if, in the case of human beings, one will die or pay taxes. An insurance company may wish to mitigate their risk by using stats formally and rigorously to predict how likely a particular group of driver is to have serious expensive car accidents - and adjust premiums appropriately. An employer may wish to use statistics informally to predict how likely a particular employee is to become a parent and quit or downsize a career, and adjust advancement opportunties appropriately. Sigh again. That people use these tools has nothing to do with where any particular datum may or may not fall. Its the entire point that the individual result DOESN'T come into play and isn't known. Because it CAN'T be known. But still you want a good guess. Saying people who smoke for decades are more likely to get lung cancer than those who don't - does not mean we need to be treating every smoker over 50 for lung cancer. Sigh again. Everybody knows this, they just refuse to acknowledge that they know it.

iVillage Member
Registered: 05-08-2003
Tue, 05-13-2003 - 5:29pm
Statistics are useful to give an idea about the "average." Sure, if you poll six graders in one class, the heights will be all over the place but if you poll six graders in the entire US -- then the heights will be more or less distributed like a bell shape with the peak of the bell at 5'4".

Another example is that the statistical chance of having a girl is 50% Now my family has 3 girls and 1 boy, and my friend comes from a family with 6 girls and 1 boy. That doesn't mean that the statistics are wrong... just that if you took a hundred (or a larger number) women who are pregnant then about 50% of them will have a girl.

Having said that,

If I hear that there is a 20% chance that this medicine will work for me, I hear it as a statistic...

it may work 100% for me or 0% because I'm an individual but that doesn't make the statistics wrong.

Pages