I think you can make do with a single array of 625 doubles (or floats) for the mean of each sample.
Calculation of the mean value for a given sample of 25,600 random numbers can be done on the fly, the 25,600 numbers need not be saved.
For B: Find the mean of the means.
This could be done on the fly as well.
For C: Find the standard deviation of the means.
Finally, some data storage is called for. I think you will need to save all the values of the means before you can find the standard deviation among them.
EDIT: I see your other thread. You don't need to generate float values for the random numbers! The float part happens when you calculate the average.
1 2 3 4 5 6
int total = 0;
for(unsigned i=1; i<=25600; ++i)
total += 1 + rand()%2;// adding up integers
float average = (float)total/25600.0f;// here comes the 1.43, etc. values