## errors propagation

Hi all,
I hope you don't bother if this is a too much crude thread...but hope someone can help me with this.

so, I 'am trying to calculate a confidence interval.

say P the total population from which a sample of N elements has been extracted.

each N has been evaluated with a series of K replicates. I have an average and a confidence interval based on the t-student value for (K-1 degrees of freedom, alfa = 0.05) for each of the N.

so I have N means and N intervals.

now, how can I use the intervals in the calculation of the grand mean of the N values.

I was thinking about the following two way.

1) to calculate a relative error &#37; form each of the N values and associate the average % to the mean of the N values.

2) forget the intervals associated to the N values and recalculate a new interval based upon N-1 degrees of freedom.

ps.
at the end, I have to extend the average N to the total population P.

thank you very much
I hope to have been sufficiently clear.