Standard deviation of one sample as an estimate for another sample's st. dev I have a set of data replicated three times and I have calculated the corresponding standard deviation. Is it correct if I use this standard deviation for another set of similar data?
I know that it is the best choice to calculate the standard deviation for each set of data by replication of the corresponding experiments. But sometimes, the experiments are so expensive or time consuming and can not be replicated practically; so I need to extend the standard deviation of one set of data to another similar set.
 A: Before you supply any extra details on your particular problem, here is a generic answer.
You estimate standard deviation of a sample to get an idea of the standard deviation of the population from which the sample was drawn. You can have a point estimate and a confidence interval around it. You can use these to anticipate the standard deviation of another sample to be drawn from the same population. 
However, if the new sample is drawn from a different population, you cannot directly apply the knowledge from the previous sample. Nevertheless, if these two populations are somehow related (e.g. you know that the second one must have a standard deviation twice as large as the first one), you could potentially adjust the previous estimate to get an idea about the standard deviation of the new sample from the new population.
A: 
I have a set of data replicated three times and I have calculated the corresponding standard deviation. Is it correct if I use this standard deviation for another set of similar data?

If you're prepared to make the assumption that they share the same standard deviation -- but it should be explicit that you're making this assumption.
If you expect other people to accept the results that follow from that assumption, you'll probably need some kind of convincing argument that the population standard deviations should be the same.
[In some cases, as you suggest, you may have little alternative. However, more details might just possibly lead to other suggestions. In some cases, for example, it may be possible to provide a bound on the standard deviation that could then provide a bound on the size of a confidence interval or a p-value]
