Sample size can be determined when you also specify the design of your experiment, the level of significance you wish to achieve and an assumption of the effect size of the phenomenon you are seeking to measure.
A common use of sample size calculations is to reduce the risk of Type II errors (false negatives) to an acceptable level (e.g. 20%). Have a look at this website: http://www.biomath.info/power/index.htm.
sample size depends upon nature of the size if it is macro level he sample will be big and micro-level the sample will be small. Generally in social science 30% sample is to bee taken. I personally believe we can justify the small as per analyzing capacity.