With the waning of the coronadoom panic, and our elites unable to juice sufficient consternation over Russia, they will have to turn to other objects to monger their fear. Global cooling, a.k.a. global warming, a.k.a. climate change is a natural.
That being so, it must be that global warming will be said to cause or to contribute to every ill.
Now you and I have seen many such cases like the paper below over the years. How these kinds of papers are generated is of keen interest, as they demonstrate the deep importance of the vast over-certainties generated by classical statistical methods. I realize discussions on their replacement are arcane and difficult, but they are necessary.
Let’s start with the propaganda headline: “Hotter nights increase risk of death from heart disease for men in early 60s”.
Speaking of heat, following the hot trend in woke advertising, the accompanying article pictures a black man. In this case grey bearded, sleeping, and, oddly, smiling—a nice dream of his deadly global-warming induced heart attack?
The peer-reviewed paper on which the headline was based in “Warmer summer nocturnal surface air temperatures and cardiovascular disease death risk: a population-based study” by Majeed and Floras in BMJ: Open.
They are concerned with temperature: “In recent summers, some populous mid-latitude to high-latitude regions have experienced greater intensification of nocturnal than daytime heat, with consequent adverse effects on human health.”
You mean like in Texas, Florida, the entire Mediterranean, where men aren’t dying at spectacularly different rates of heart attacks?
Well, you know what causes temperature, don’t you? Yes: climate change. So this is an important paper.
They gathered minimum temperature data for three locations in June and July, one location in England, one Wales, and one in King County, Washington. They also had dust measurements (PM2.5) for Washington.
Next came cardiovascular (CVD) deaths for 2001–2015. And, for a fun reason I’ll explain below, they also collected deaths from “mental and behavioural disorders”. My favorite method is this:
Sex-specific analyses were partitioned into two age groups: 60–64 years and 65–69 years. We elected to exclude from analysis younger adults, due to their lower CVD event rates and older adults, since in England the cause of death of individuals ≥75 years of age is likely to be misclassified, due to their higher prevalence of comorbid conditions.
Misclassified. Uh huh. Except for covid, of course. Do we think they were tempted to examine the 75+ group anyway, and left their results out of the write up for some reason?
Either way, this analysis is over a very narrow range of ages.
Now you’d think they’d tie number of CVD deaths to temperature. But no. They looked instead of deviations of minimum temperature from a baseline. Such deviations are commonly called “anomalies”, though they aren’t anomalous.
Here’s where it gets amusing (with my emphasis and paragraphification):
CVD mortality rates were found to be autocorrelated…Additionally, the outcome variable’s variance was much greater than its mean, leading to overdispersion of data. Moreover, a previous study showed that the incidence of mental health and behavioural distress in England and Wales has both increased over time…To address these issues in our models, we used negative binomial regression with autocorrelated residuals of order one to assess the association between sex-specific and age-specific CVD mortality rates to summer nocturnal SAT [surface air temperature] for England and Wales from 2001 to 2015, while controlling for each of mental health and behaviour mortality rates, an increase or decrease in CVD mortality rates with respect to the annual calendar year (i.e. trend) and the summer month as our covariates.
For King County, we used quasi-Poisson to assess all associations, while controlling for each of PM2.5, an increase or decrease in CVD mortality rates with respect to the annual calendar year (ie, trend), and the summer month as our covariates. Findings are reported as incidence rate ratios (RR) and interpreted as change for one-unit increase of the exposure variable.
This is all screwy.
Long time readers know deaths from most causes (including CVD) have a strong periodicity, bottoming out in summer. Hence the “autocorrelation” (which does not mean CVD deaths from one month cause CVD deaths in the next month). But why on earth are they “controlling” for deaths by lunacy? Why not deaths for stroke, or cancer? Why model any deaths at all beyond CVD? Recall statistical “control” has no relation to actual physical control.
And what’s all this about “controlling” CVD deaths for increases or decreases in CVD, when the object of the study is increases or decreases of CVD!
There was also no good reason to model Washington with PM2.5; at least, a model without it that matched the Brit one should have been presented concurrently.
Anyway, after all that, here is the stunning “finding”:
[A] +1°C anomalous summer nocturnal SAT associated significantly with an increased risk of summer CVD mortality rates among men aged 60–64 (adjusted RR 1.031; 95% CI 1.003 to 1.059) but not in those aged 65–69 years (adjusted RR 0.999; 95% CI, 0.976 to 1.021), nor in adult women in either age group.
Since it’s only younger males in July, and not females, and since the “effect” is so tiny, and since all the “controlling” variables make no sense, and since this is confirmed by wee p-values (confidence intervals are equivalent), this is clearly statistical nonsense.
They are unable to consider over-certainty, saying instead that “The non-significant trends observed for the older men in the present analysis and in these previous reports may reflect resilient survivor bias”. Uh huh.
Here’s another fun one: “Although we cannot infer causality from our models, our age- and sex-specific analyses nonetheless represent a novel contribution to the present literature.”
You’d be hard pressed to find any paper that doesn’t claim to be novel. Yet all will admit that correlation isn’t causation wink wink wink wink.
Being good virtue signalers, they couldn’t resist closing with this: “Considering the growing likelihood of extreme summers in Western USA and UK, our results invite preventive population health initiatives and novel urban policies aimed at reducing future risk of CVD events.”
Again, if this analysis (like all analyses) was cast in predictive form, as I am always preaching, these errors would not be so blatant.
Buy my new book and learn to argue against the regime: Everything You Believe Is Wrong.