Using externally calculated st.dev. in X by Y

[#1]

Hi everyone,

i occasionally get datasets, where there is no raw data available, just the mean, st.dev. and number of replicates. Is there a way to use this available data to run a t-test, for instance? A quick platform launch is included to show what I want to do.

I was thinking, that maybe constructing a simulated distribution (let's assume it's normal) with the given parameters and counts could be the way to go, but I I'm not sure how to do that. Does anyone have an idea? A scripted solution would be good, i could adapt that.

Have a great day!

Kanstantsin

[#2]

You can use --> Sample Index --> Teaching Scripts for a t-test with summary statistics:

But may be not appropriate for many rows.

[#3]

Hi Georg, thanks for this! Didn't know this option existed! Doesn't scale too greatly unfortunately.

The solution that Mark gave is what I was looking for! And, yeah, that's not a dataset with high effects...

[#4 · Accepted solution]

You might also be able to use this solution.

You would have to run very many bootstrap samples to approach inference by resampling.

[#5]

Hi Mark,

thanks a lot! yes, this worked rather nicely, and so did the concatenation for categories that Ron mentions in his reply.

[#6]

I would probably use this graph to understand the data, mean on y and stddev as interval:

Resampling can be done of course, but will not lead to a good assessment with small differences.