0

I tried to find this question but what I've found provides answers for normal distributions and large sample sizes.

I'm analyzing some data from an experiment with different groups. Each group has around 20 subjects.

When data is normal and homogeneity of variances is met I plot mean scores and standard error of the mean.

When assumptions of normality and homogeneity of variances are not met I would like to plot medians but I don't know what would be the Standard error equivalent for small non normal samples.

Any ideas? Thanks in advance

Gonzalo Lerner
  • 21
  • 2
  • It depends on the distribution from which the samples were drawn. – Glen_b Nov 27 '18 at 13:43
  • To extend glen_b's comment, it depends on the exact data. Hopefully, you know enough about the data to hypothesize a distribution from which they arise. For example, the data could be reaction times informing us they're bounded at zero, that's some information to start with hypothesizing. – Heteroskedastic Jim Nov 27 '18 at 14:05
  • Have you looked at https://stats.stackexchange.com/questions/45124? That answers the question about the sampling variance of medians. But perhaps what you really need is a *nonparametric confidence interval* for the median; thus, some editing of this question to clarify your needs would be welcome. – whuber Nov 27 '18 at 16:16

0 Answers0