# Stork Example

Six Sigma – iSixSigma Forums Old Forums General Stork Example

Viewing 9 posts - 1 through 9 (of 9 total)
• Author
Posts
• #28811

Trent travis
Member

Hey Troops
Can someone tell me the stork story as I need to teach Correlation and I need a good story to hi-light the fact that though relationships may be strongly correlated, in truth they simply mean there is no cause and effect relationship.

Hope someone can help

0
#72334

Joy
Participant

The example comes from the book, Statistics for Experimenters by Box, Hunter & Hunter, p. 8.  The population of Oldenburg at the end of each of 7 years is plotted against the number of storks observed in that year.  There is a very strong positive correlation in the plot but few would be lead to believe that the increase in the number or observed storks CAUSED the increase in population.
I hope this is what you needed.
Joy

0
#72335

Erik L
Participant

The other standard example that I’ve seen is to look at ice cream sales and shark attacks.  While they show a strong correlation it does not imply causation.
Regards,
Erik

0
#72344

Luke Ng
Participant

Hi,Another interesting example would be the strong positive correlation between “the severity of damage” and “the number of firemen”. “More firemen” does not cause “more damage”.Cheers,Luke

0
#72356

Ed Blackman
Participant

My favorite correlation example is the relationship between country radio stations and suicides. The greater the number of country radio stations in a given population area (a city), the greater the number of suicides. A rather high positive correlation exists. However, the conclusion that country radio causes people to end their existence is erroneous. It turns out that the larger the population in a city is, the more country stations they tend to have, and of course, more suicides.

0
#72414

Trent travis
Member

Thanx Troops

Some good explanations.  Keep em coming if you have any more

Cheers
T

0
#72425

Gabriel Braun
Participant

Nice examples, but be careful. Don’t kill the relationship between correlation and casue-efect too fast!. When there is a strong correlation between x and y, it is likely that there is a cause-effect chain involved. What you don’t know is which is the cause and which the effect. Usually, at least one of the variables (x or y)is an effect, but you don’t know which one. If x is an effect, it is possible that y is the cause or another effect from a common cause. So you may not have the cause among your variables x and y! Examples: The cause of increase in “icecream sales” is “higher temp”. The cause of increase in “shark attacks” is “more swimmers”, due to “high temps”. So the “summer” is a common cause for both effects. In the examples of the storks and the population, of course that the increase in “observations” does not increase the “population”, but the increase in “population” (cause) does increase the “observations” (effect). A strong correlation is a first step towards a cause-effect analisys. But you must be open minded and must not begin that analysis with assumptions. The only information you have is that both variables are correlated and then they have a common cause. You know that at least one of the variables is an effect. One of the variables might be that common cause itself. But the correlation chart can’t tell you which one is an effect and which one (if any) is a cause. The cause-effect analysis is needed for that.

0
#72497

Bobby
Participant

An example I have always liked shows a correlation between the number of shark attacks and ice cream sales in dollars in California.  As one increases so does the other.
Reflecting on this, it seems absurd to assume that an increase/decrease in ice cream sales will have a corresponding impact on the number of shark attacks.  And to assume people will want to eat more ice cream when the number of shark attacks increase is just as absurd.  The actual connection is the temperature.
Hope this helps.

0
#72522

Kevin Hankins
Participant

Correlation and regression are rich areas which need much applied R&D before they can be broadly and commonly applied in SS programs.
For instance, I have seen none of the available SS materials explain the difference in shape of the prediction intervals between correlation and regression.  Or that the univariate spec limits on which SS is based is an improper shape for any real product having multiple responses(rectangular cuboid vs proper ellipsoid).
Any MBB interested in helping me in this applied R&D, please email me.
Kevin HankinsMaster of [email protected]

0
Viewing 9 posts - 1 through 9 (of 9 total)

The forum ‘General’ is closed to new topics and replies.