Look at the standard error, page-62

This really isn’t too relevant to what I posted. I agree that just looking at overlapping SE is very conservative. On the other hand if you use the 95% confidence intervals (CI) then you have the problem where you reject groups as not being significantly different that actually are different.

Having said this you are on safe ground that if you find that if the SEs overlap that there is no significant difference between the groups. Yes you will have more Type I errors by doing this, but in this case it doesn’t matter.

For those that are interested in this sort of technical question apparently the answer (if the SEs are approximately the same size) is to use 83% of the CI value [1]. I have to say I didn’t know this before now, but I don’t claim to be a statistics expert - I know just enough to know how to spot the obvious and know when I need to call in an expert.

1. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC524673/