[R] Means do not tally

[Ben Bolker]

Ruijie <breakaway8 <at> gmail.com> writes:

> In my experiment, I expose 10 subjects to 24 different conditions of
> stimuli. Each condition is exposed to the same subject 3x.
> This would make each subject have 24x3=72 data points. All the subjects
> combined would have 72x10=720 data points with each condition having 30
> datapoints.
> 
> To find the grand average of each condition, I find the average of all the
> datapoints for a given condition.

 [snip]

> Since I computed the average of each condition per subject, theoretically,
> if i average the average of each condition per subject across all subjects,
> the result would be the same as the grand average.
> 
> However, this is not the case when I use R or Excel. regardless of functions
> used. Anyone have any thoughts?

  Nothing obvious springs to mind. We need a reproducible example.
How different are the means computed in different ways?  Is the
difference attributable to round-off error?