Lying with numbers

99% of all numbers provided to support an argument — be it in scientific or popular media — have no value whatsoever unless they come with proper mention of at least the confidence intervals, the size of the sample and its properties, or other means of making reliability comparable such as standard deviation.

[w] [w] [w]

2 Responses to “Lying with numbers”

  1. rattus says:

    Maybe I should have added that, however, these numbers sometimes are good for entertainment. But then, maybe this was obvious ;-)

  2. rattus says:

    The latest “Bad Science” column by Ben Goldacre explains the role of confidence intervals in statistical studies, and gives an example from the British media of how reporters ignore them. I’ve discussed this issue previously (see, for instance, the Resource below), but mainly in the context of public opinion polling. However, any study that uses sampling to draw a conclusion about a population will have a confidence level and margin of error. Fallacy Files, 2011-08-20