I have a batch of material (25kg limestone) which contains low concentrations of several elements. 100 g samples are taken from this batch and undergo a process which may lead to the concentrations of these elements to increase or decrease (though more likely to increase). Following the process, 20 mg sub-samples are taken from the 100 g samples and are analysed to see if there is any change in the concentrations.

I need to use a statistical test which enables me to say with a given certainty whether or not there has been increase in the concentrations following the process. This can be done using the student’s t-test, which I have used before.

However, I am unsure how many samples and sub-samples I would need to take and analyse, given the size ratios given above. If the number of samples is dictated by the desired probability (e.g. 95 %, 99 %), how do i go about choosing the correct probability for these particular experiments?