Fallacy - the Counterfeit of Argument

[SlideRule]

Scientific American - Quick & Dirty Tips
"A few weeks ago, our family took a ferry across the English Channel. On the way from England to France, the ferry was extremely crowded, and the crossing took about 75 minutes. On the way back from France to England, there was hardly anyone on the ferry, and the captain announced that the crossing would take 95 minutes.
My 5-year old, wondering why the crossing was going to take longer on the way back, came up with this hypothesis:
“Maybe when there are less people, the boat goes slow; but when there are more people the boat goes faster.”
All the facts seemed to support her, and while this could be possible, I’ll hope you’ll agree that it wasn’t the most likely explanation. My 5-year-old had fallen prey to a classic statistical fallacy: correlation is not causation.
This phrase is so well known, that even people who don’t know anything about statistics often know this to be true. But the thing is, sometimes in science correlation is all you’ve got.
Chopping Latin
The official name for this type of logical fallacy is 'Cum hoc ergo propter hoc,' or 'with this, therefore because of this.' According to my daughter’s reasoning, after fewer people got on the ferry, the trip took longer. Therefore the trip took longer because fewer people got on the ferry.
It’s easy to see the problem with that logic in these examples:
'After I washed my car, it rained. Therefore washing my car causes rain.'
'When I got in the bath tub, the phone rang. Therefore getting in the bath will lead to the phone ringing.'
'We won our baseball game when I was wearing these socks, so it must be the lucky socks that caused our win.'”

Correlation does not imply causation.

BEST!
SlideRule

[Thomas Okken]

(08-31-2019 02:27 PM)SlideRule Wrote:  correlation is not causation

Almost invariably, when that line gets quoted, it is abused.

Here's my take: of course correlation implies causation! The only alternative is coincidence. In the example of the ferry ride, the data set is so small that coincidence is basically guaranteed. You can always fit a straight line perfectly to a data set consisting of two points. But the greater your data set, the less likely coincidence becomes. Which is why initial studies that show interesting results should always be followed up by more, and larger, studies.

I think what people abusing the "correlation is not causation" line often mean to say is that correlation tells you nothing about the direction of causation, as in the famous example of the correlation between owning a gold watch and being obese in 19th century England. But if A and B are correlated, and the data set is large enough to rule out coincidence, then there most definitely is a causal relationship. Whether that relationship is A causing B, or B causing A, or A and B both being caused by some other variable C, that is a different question.

[SlideRule]

(08-31-2019 03:53 PM)Thomas Okken Wrote:  

(08-31-2019 02:27 PM)SlideRule Wrote:  correlation is not causation

Almost invariably, when that line gets quoted, it is abused.

What quotation? My closing statement was Correlation does not imply causation, sans quotes. As to the counter-statement, I offer no argument, pro or con.

“Correlation Does Not Imply Cause” - But What Does That Actually Mean?

"It seems pretty self-explanatory, but it's not always easy to understand exactly that this phrase means until you examine it carefully. First of all, it is important to understand what a correlation is and what a causation is. A correlation is a mutual relationship or a connection between two variables. Causation is the relationship between cause and effect. So, when a cause results in an effect, that's a causation. In other words, correlation between two events or variables simply indicates that a relationship exists, whereas causation is more specific and says that one event actually causes the other.

When we say that correlation does not imply cause, we mean that just because you can see a connection or a mutual relationship between two variables, it doesn't necessarily mean that one causes the other. Of course, it might be the case that one event or variable causes the other, but we can't know that by looking at the correlation alone. More research would be necessary before that conclusion could be reached."
source: Why Correlation does not Imply Causation in Statistics

BEST!
SlideRule

[toml_12953]

(08-31-2019 02:27 PM)SlideRule Wrote:  Scientific American - Quick & Dirty Tips
"A few weeks ago, our family took a ferry across the English Channel. On the way from England to France, the ferry was extremely crowded, and the crossing took about 75 minutes. On the way back from France to England, there was hardly anyone on the ferry, and the captain announced that the crossing would take 95 minutes.
My 5-year old, wondering why the crossing was going to take longer on the way back, came up with this hypothesis:
“Maybe when there are less people, the boat goes slow; but when there are more people the boat goes faster.”
All the facts seemed to support her, and while this could be possible, I’ll hope you’ll agree that it wasn’t the most likely explanation. My 5-year-old had fallen prey to a classic statistical fallacy: correlation is not causation.
This phrase is so well known, that even people who don’t know anything about statistics often know this to be true. But the thing is, sometimes in science correlation is all you’ve got.
Chopping Latin
The official name for this type of logical fallacy is 'Cum hoc ergo propter hoc,' or 'with this, therefore because of this.' According to my daughter’s reasoning, after fewer people got on the ferry, the trip took longer. Therefore the trip took longer because fewer people got on the ferry.
It’s easy to see the problem with that logic in these examples:
'After I washed my car, it rained. Therefore washing my car causes rain.'
'When I got in the bath tub, the phone rang. Therefore getting in the bath will lead to the phone ringing.'
'We won our baseball game when I was wearing these socks, so it must be the lucky socks that caused our win.'”

Correlation does not imply causation.

BEST!
SlideRule

Oh yeah? I believe Betadyne solution causes blood loss. Every time I go to the Red Cross and get that stuff swabbed on my arm, I lose a pint of blood!

[cdmackay]

(08-31-2019 03:53 PM)Thomas Okken Wrote:  […] or A and B both being caused by some other variable C, that is a different question.

Yes, that's the vital bit, that is often missed…

e.g. the data shows that people living in area X are more likely to develop disease Y.

Is that because there's something about area X that is causing disease Y, e.g. some chemical plant emitting noxious gas?

Or perhaps area X has a higher-proportion of lower-income people, lacks facilities, and has a lower level of public health in general, and disease Y is a consequence of that?

Or even a combination of the two, and many other factors.

Stats is (relatively) easy; interpreting stats correctly, not so easy.

[grsbanks]

"There are lies, damn lies and statistics."

[JimP]

(08-31-2019 02:27 PM)SlideRule Wrote:  Correlation does not imply causation.

Sadly, there are a lot of scientists (many of them physicians) who believe they have discovered a cause for some kind of effect, by citing a correlation. Upon publication, this can be like a snowball rolling downhill. Years later, often with much money spent, and much money made, this correlation becomes part of everyday life (or worse, "treatment guidelines"!) It is indeed the bete noir of good methodology, for once embedded in accepted literature, the work required to remove it is not only astronomically hard, it is also met with severe resistance from those whose financial, or control positions might be threatened as a result.