http://co2coalition.org/2020/06/16/climate-statistics-101/
This is the slide show and 20-minute talk that Representatives Alexandria Ocasio-Cortez and Chellie Pingree tried to censor at the LibertyCon 2020 conference in Washington, D.C. After Dr. Rossiter gave a climate talk at LibertyCon 2019, they wrote to sponsors of the event, such as Google and Facebook, and asked them not to fund any event with an appearance by “climate deniers” from the CO2 Coalition. See http://co2coalition.org/2019/01/30/representatives-ocasio-cortez-and-pingree-and-climate-change-debate/
LibertyCon indeed lost some sponsorship, but because of its commitment to the free exchange of ideas still invited Dr. Rossiter back to speak in 2020. This is the talk he had prepared, before the coronavirus crisis forced the cancellation of the conference.
As background to this topic, we suggest that you watch the CO2 Coalition’s “CO2-Minute” video, “Carbon Dioxide: Part of a Greener Future,” at https://co2coalition.org/studies-resources/video-and-media/.
Now, on to the talk! (You can also download and distribute the slides themselves in a PowerPoint file at: http://co2coalition.org/wp-content/uploads/2020/06/LibertyCon-Rossiter-Presentation-final_6-16-20.pptx)
I’m Caleb Rossiter, executive director of the CO2 Coalition of climate scientists, and a former statistics professor. Welcome to Climate Statistics 101, which shows how to test hypotheses about the impact of emissions of greenhouse gases like CO2.
Statistics uses logic and probability to test for causation, for whether one thing affects another. We take nothing on faith, everything on proof. Only in the law school do they teach ad homimen arguments – attacking or praising the messengers. Scholars just analyze their message.
This is life! It’s called the Normal Distribution or Bell Curve. It shows how far away from the average most physical and statistical things are. Things like people’s heights or the number of hurricanes in a decade.
We use the Normal Distribution to test the null hypothesis, the claim that there is no “statistically significant” difference between the average and what we actually observe. Most of the time, 68 percent of the time, observations are close to the average, within one standard deviation – the average distance of the data from the average itself. As you move farther from the average, you get less of whatever it is you are counting. There are a lot more six-foot guys than seven-foot guys.
This formula, derived from our mathematics and amazingly confirmed in nature, determines the height of the Normal curve at every point. It tells us just how often what we observe will be, simply by chance, a certain number of standard deviations away from the average.
This “Z-table” tells you exactly, to the third decimal place, how likely it is that an observation happened by chance. When we run an experiment, we only reject the null hypothesis, and say there is a statistically significant correlation, if the outcome would happened anyway one time out of 20, or five percent of the time. That makes us 95 percent sure that the two variables are correlated, or move together.
Now, correlation is not necessarily causation. Life is not bivariate – based on just the two things you are measuring. Unless you can randomly assign subjects, life is multivariate, with other variables causing changes too. This is the most common error in public policy. In Latin it’s called post hoc ergo propter hoc: this thing happened after that thing, so it was caused by it. Here’s an example.
Full slide show here: