#Scatter plot correlation funny series

#Scatter plot correlation funny series

The absurdity of doing this is a little less transparent when we’re dealing with time series (values collected over time). But this is exactly analogous to what the author of the image above is doing by including the correlation coefficient. For example, if you wanted to know the average height of people in Michigan, but you only collected data from people 10 and younger, the average height of your sample would not be a good estimate of the height of the overall population. To be able to do this, your sample must be a good representative of the population, otherwise your sample statistic will not be a good approximation of the population statistic. In the case of time series, we’re using data from a short interval of time to infer what would happen if the time series went on forever. When we calculate statistics that summarize values of a variable (like the mean or standard deviation) or the relationship between two variables (correlation), we’re using a sample of the data to draw conclusions about the population. It’s that including the correlation coefficient on the plot is misleading and disingenuous, intentionally or not. The problem I have with images like this is not the message that one needs to be careful when using statistics (which is true), or that lots of seemingly unrelated things are somewhat correlated with each other (also true). Recently there have been several posts on the interwebs supposedly demonstrating spurious correlations between different things.