nanmean giving nansum output

function 'nanmean' is behaving like nansum and I don't know why. the help command (image below) is also referencing nansum's functionality.

any ideas? i recently updated to 2022b..

Answer 1

[the cyclist]

That's not the help I get for nanmean (in the Statistics and Machine Learning Toolbox). Looks like there is a nanmean in the Financial Toolbox as well, but this seems like odd behavior.

Is it possible that nanmean points to a user-defined function instead? What output do you get from

which nanmean -all

[Ralph Andrews]

ahh yes, thank you, that's exactly what i needed to see.

looks like i had a few competing nanmean functions and i have now moved the toolbox/stats one up in the path :)

Answer 2

[Matt J]

function 'nanmean' is behaving like nansum

In what way? Example?

i recently updated to 2022b..

If so, why use nanmean at all? As the documentation told you, it's obsolete. Instead, use

Y=mean(X,dim,'omitnan')

[Ralph Andrews]

more typing init

[Matt J]

Then wrap it in a shorter function...

>> nanavg([1,2,3, nan, nan])

ans =

2

function out=nanavg(varargin)

out=mean(varargin{:},'omitnan');

end

Or, abbreviate the flag:

x=[1,2,3, nan, nan];

mean(x,'o')

ans = 2

[the cyclist]

@Ralph Andrews, I would at least consider using the (officially recommended) syntax that @Matt J shared.

I don't know how experienced a user you are, but my perspective is that a little bit of extra typing is always worth some extra robustness and code clarity. (I would definitely not use 'o' instead of 'omitnan', though!)

At the very least, I would suggest reading the version history of nanmean, and understand the reasons that it is not recommended. (But, there are no plans to remove that function, so you are safe in that sense.)

[Ralph Andrews]

that's cool i didn't think about wrapping it in a function, i definitely find it a lot more fiddly and time consuming to insert 'omitnan' flags instead of just 'nanmean'. and that's interesting to see why nanmean is not preferred, but i think for my usages nanmean does the job just fine :)