Misspecification happens when an important variable is omitted, or a nonlinear relationship is modeled as linear, and can have an assortment of pathologies associated with it, but in general, regression is not a good way to sort out causality anyway, and the fact that your independent variable is strongly correlated with your dependent variable is about all that you can learn with this tool.

The problem here is that the independent variable is time series and serially correlated and trending. If you regress one trending variable on another or others you automatically explain nearly all the variance in the independent variable – doesn’t even matter if they are trending the same way (just get negative correlation). Rsquared values of .98 are pretty typical of regressions of nonstationary time series.

While whole books and graduate courses are devoted to figuring out how to deal with this problem, a quick and dirty one is to detrend the data and then look for correlation. So instead of regressing the actual incidence you could take the change in incidence as your y variable, and regress on the change in x, now the constant term will capture the trend and only variation from trend will be modeled. Without even doing the math you can see that the deviations are inversely correlated, the positive deviations from trend for breast cancer occur when the growth in abortion ticks down and visa versa.

E

“From the three you then use one
To make ten ones…
(And you know why four plus minus one
Plus ten is fourteen minus one?
‘Cause addition is commutative, right.)
And so you have thirteen tens,
And you take away seven,
And that leaves five…

Well, six actually.
But the idea is the important thing.

Now go back to the hundreds place,
And you’re left with two.
And you take away one from two,
And that leaves…?

Everybody get one?
Not bad for the first day!

Hooray for new math,
New-hoo-hoo-math,
It won’t do you a bit of good to review math.
It’s so simple,
So very simple,
That only a child can do it! “

That doesn’t explain the disparity in the number of conservative op-ed columnists they publish vs the much smaller number of liberals; heck, even the relatively smaller number of center-left if you want to be charitable.

There are clearly liberals among their readership also. Not all of us vote for Rubber Chicken (R).

The same bottom-line argument is made as an excuse for the number of conservative pundits on cable and broadcast tv. And, again, it’s a false argument given that the most watched shows present a center-left point of view. (That “center-left” distinction is important because even among those pundits actual “liberal” perspectives are few and far between.)

Now where’s my Tribune Column gig?

Fewer children, or fewer children breastfed, and scientists see an increase in the risk of breast cancer.

Given the prevalence of formula-fed babies in developed countries, a lack of nursing is a much more likely cause (or one of many causes) for the increased rates of breast cancer we’re seeing. (Use of pesticides, hormones, antibiotics, and other potential carcinogens in our modern industrialized agriculture could be another.)