[quote]smh_23 wrote:
[quote]pat wrote:
[quote]smh_23 wrote:
[quote]Da Man reloaded wrote:
[quote]smh_23 wrote:
[quote]pat wrote:
lol on the graph fail.
The peak closing of clinics correlates very well with the teen pregnancy drops. You illustrated kneedraggers point to the T.[/quote]
What?
From 1990 to 1996, the number of PP clinics in the United States rose from about 880 to 936. Concurrently, the teen pregnancy rate fell from 120 to 100 (per thousand).
Again, the number of PP clinics rose and fell and rose and fell and this is not reflected in the teen pregnancy rate even slightly.
The study cited in the OP is about as shoddy as this kind of thing can get.[/quote]
overall trend is down. curve fit that bitch.[/quote]
Not in the years that correspond with the teen pregnancy-rate graph. The overall trend is effectively static. A net change of essentially nothing.
And even if the two were both trending down overall, this would say exactly nothing about correlation. If statistically significant, sustained peaks and valleys don’t match up, then you’re fighting an uphill battle trying to tell anyone that correlation is present.
Alright, I’m done arguing over statistics and graphs. If you can look at those two data visualizations and see correlation, then you need to reevaluate your understanding of basic mathematics, or the bias-blindness which allows you to ignore your intelligence–because I know for a fact that I’m arguing with intelligent people here–and instead conclude what you’d like to conclude. Also, I consider a math debate to be over once Dr. Matt weighs in on it, which he has.
Next time, let’s argue about the quadratic equation.[/quote]
Incorrect. You can make a correlation with anything that has a commonality with another. In this case the common thread is a downward trend. To show that the two data sets do not correlate you have to show they have nothing in common.
The correlation may be low, but it’s still there based on that very small amount of data which objectively we don’t even know to be accurate.[/quote]
You are incorrect. A correlation is a statistical measure that indicates the extent to which two variables co-fluctuate. Direct correlation inplies both co-rise and co-fall. Absent one of these, and assuming that the rises and falls were statistically significant–which they were–you’re grasping at straws that aren’t there when you start averring correlation.
The fluctuations of the PP clinic set of data have exactly no analogous fluctuations in the teen-pregnancy-rate set of data. If the two were correlated, a statistically significant and sustained rise in one of them would coincide with a statistically significant and sustained rise in the other. It doesn’t.
If one variable trends up and down alike, in effectively equal doses over a given amount of time, and the other trends only and almost perfectly uniformly down over that same given amount of time, they are not correlated over that given amount of time.
I urge you to consider this: you are essentially defining correlation as “well they both ended up below where they started, no matter how they got there and no matter that one fell by a statistically minute downtick while the other nosedived.” This is meaningless and not correlation. Want to know why? Because, by that standard, any two variables, any two sets of data no matter how disparate or different or baldly unrelated, are correlated, either directly or inversely. Number of left-handed rodeo-clowns and number of abortions in the United States? Correlated. Average penis size of Senegalese ranchers and average height of German street performers since 1900? Correlated. All correlated, either directly or inversely. Cause look–one ended up, the other ended up. One ended down, the other ended down. One ended up, the other ended down. Correlation.
No. It has been explained again and again that those two graphs do not betray the rudest outline of a correlative relationship. This is manifestly clear in that their fluctuations do not match up, match-up of fluctuations being the very definition of correlation.
“The correlation may be low”–when you said they were correlated, you were saying that the correlation was high. If the fluctuations do not match up, we say that they are not correlated. If they do match up, we say that they are correlated.[/quote]
Sigh…
The fluctuations do not have to match up 1:1 for there to be a correlation. One can fluctuate and the other not so long as they are trending the same way.
For instance, if you had leukemia and were trending towards remission, your white blood cell count can still fluctuate into a range indicative of untreated leukemia, but so long as the overall trend is downward, it can be correlated to overall remission, even with fluctuations.
Fluctuations in the Sun’s temperature does not correlate with the Earth temperature fluctuations 1:1, but the source of most of the Earth’s heat is still the sun.
One actually would not expect the data not to fluctuate over time. Even with the fluctuations the data is still trending downwards as the other is trending downwards.