Hi,

First, am I correct in saying the following?

When inferring from 95% confidence intervals whether the difference of two means is significant on a 5% level, two steps are necessary:

- multiple the CI-width by Square Root of 2 (=1.414...)

- neither of the CIs must overlap the mean of the other group

If the latter condition is met, p < .05, otherwise p >= .05.

Second, if yes, where does Square Root of 2 come from? I am aware of the Loftus and Masson (1994) paper: Article Using confidence intervals in within-subject design

I am looking forward to your answers!