 Okay, so let's look at the residuals. That is to say what's left over in the data when we remove this regression line. Now we can use the regression model tool here. The regression that we've done is where our independent variable is year. Our dependent variable is temperature. And so we can run that regression here. And that gives us two things. First of all, it tells us the autocorrelation in the residuals, which as we can see is pretty small. It's minus 0.11. And if we look up the statistical significance of a negative correlation of 0.11 with on the order of 100 degrees of freedom, we'll find that it's statistically insignificant. So that means that in this particular case, it doesn't look like we have to worry about the added caveats associated with autocorrelation, when our residuals do not look like uncorrelated white noise, but instead have this low frequency structure. Now we can actually plot those residuals, and I'll make a new plot here. I go down to model residuals, and so I'm going to plot the residuals as a function of year. We no longer need a trend line here, and that's what we have. So when we remove the trend, we accounted for a statistically significant trend, and when we removed that trend, this is what was left over. These are the residuals, and they look pretty much like Gaussian random white noise, which is good. That means that the results of our regression are basically sound. We fulfilled the basic underlying assumptions that what's left over after we account for the significant trend in the data looks random.