How Not to Lie with Statistics by Darrell Huff

Be Careful of Half Truths and Statisticulations

Something New I Learned

I didn’t know that the word “average” could mean two different things. I always thought that the word referred to the mean and not the median. But Huff writes that it could refer to the mean, median, or mode, and authors usually pick one of them that suits their purpose the best and represent it as the average.

It is important to report the range and not just the single number statistic. Huff talks about the Gesell’s norms, which put many parents into concerns for their children who developed later than the age, and not a range of ages, indicated by the Gesell’s norm. Or, the two close IQ scores between two people, such as one of which is a 99 and another 102, doesn’t determine who is smarter than the other. A more meaningful indicator is the range of values, which include values that are likely to occur with 50 percent chance. Only a range of values can take the error of measurement into account to yield better accuracy.

It is important to be careful of comparisons represented by pictures. People are prone to comparing areas when they look at pictures. They usually don’t measure the respective heights or lengths of different figures to determine how much greater or smaller a figure is than another figure. Therefore, pictures tend to exaggerate because the ratio of the difference is squared when people compare different areas, not lengths or heights.

Good Reminders

Statistics is much an art as it is science. Statisticulation is a word I had never heard before, but I have definitely heard about the idea which it refers to before. It represents the fact that various calculations can be made to suit different interests even when those calculations are made without any fabrication. To be honest with statistics, it is simply not enough that one does not lie but that one does not pick the wrong method. Huff illustrates how a statistician can avoid lying but still be dishonest. In a map that represents the federal spending by coloring the states whose total incomes accumulate to the spending, large western states are colored which makes an untrained eye to think that the spending was huge. However, if someone wanted other people to instinctively think that the spending was small, it would have been equally valid to color the smaller eastern states. This demonstrates that coloring the maps to represent the spending misleads and is not a completely honest way of presenting it.

A statistician confronts an equally challenging problem when dealing with percentages and numbers. Sometimes percentages favor one party while numbers favor the other side. Huffs quotes Times as an example of good statistical reporting when it presented a graph that favors laborers overlapped with another graph that favors executives.

Never believe a comparison without knowing the nature of the two things that are being compared. Depending on the base, a figure can be said to have increased or decreased from before.

Examples from My Own Life

To be continued…