Does it make sense to look at a scatterplot of means versus standard deviations in terms of looking for outliers or changepoints? So the mean is on the x-axis and the standard deviation is on the y-axis. We would get a clusters of data indicating similar data (relatively same mean and standard deviation). Note that I don't know the sample size or individual observations from which these means and standard deviations were computed from.

  • $\begingroup$ Why not? If you weigh a person 3 times, average that and then use that average instead it should not be that much different from just choosing one weight. It is also interesting to the relation to the standard deviation of those weights as this gives you additional valuable information. $\endgroup$
    – Max Gordon
    Commented Feb 28, 2013 at 21:57
  • $\begingroup$ If, in a final analysis, you are using an aggregate measurement that has imprecision between measurements (less significant for weight, but consider blood pressure), you MUST account for that level of variation in your analysis. Inverse variance weighting is a common approach to this. $\endgroup$
    – AdamO
    Commented Mar 1, 2013 at 0:25

1 Answer 1



I assume you're examining data from a continuous outcome that is somehow clustered at a group level. An example of this might be inflation adjusted profit margins over years for several restaurants in the same zipcode, though I'm curious what your data represent.

There is nothing fundamentally instructive about the relationship between means and standard deviations unless part of a prespecified analysis plan. Instead, you can display summaries about the spread of data at each aggregate level using a box-and-whiskers plot, or a forest plot with 95% confidence intervals for the observed mean values.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.