Everyone talks about global warming. This talk is about local warming. I grabbed from the NOAA website 55 years of daily average temperature data from McGuire AFB here in NJ, and made a regression model that includes a sinusoidal cycle with a 1-year period to model seasonal fluctuations, another sinusoidal term for the 11-year solar cycle, as well as a linear trend. I consider two regression models: a least squares (LS) regression and a least absolute deviations (LAD) regression. As is well-known, LAD regressions are insensitive to outliers in the data. It turns out that the LAD model correctly identified both the period and the phase of the solar cycle, whereas the LS regression got it completely wrong. And, the amplitude solar cycle is small, so this lends confidence to the overall results beyond what one my get from a simple confidence interval. Of course, you probably want to know if my model establishes the existence of local warming in New Jersey or not. And, if so, what's the rate of warming? Come to the talk to hear the results.
Group for Research in Decision Analysis