Anscombe dataset No. 5: Simpson's paradox

A new fifth dataset has been created, using the very same independent variable as three out of the four original datasets, and is in complete concordance with the summary statistics of the original Anscombe's quartet. On this occasion the new dataset serves as an illustration of the so-called Simpson's paradox. The csv dataset provided here is comprised of the original Anscombe quartet (as per the table in F. J. Anscombe "Graphs in Statistical Analysis", The American Statistician, vol. 27, pp. 17-21 (1973) doi:10.1080/00031305.1973.10478966 ) along with a new column (y5) which constitutes an example of Simpson's paradox. Only the data corresponding to column y5 is original work. All other columns are the work of F. J. Anscombe and are included for completeness, context, and convenience. See the pdf file associated with this dataset for details as to how this new column was created.