First it is important to define a representative sample. A representative sample should have a mean very close to the entire population being sampled, i.e. you are able to account for the natural variation occurring in the population. Consequently, for more variable populations, a larger sample size should be expected in order to collect representative data.Then discuss how appropriate samples should be collected. A common strategy is random sampling, and is important to avoid any observer bias. This can be achieved through methods such as random coordinate generators during quadratic sampling or random number generators when working with numbered samples. If data can be split into different groups and you want to still get an equal number of samples from each group then stratified or pseudo-random sampling may be more appropriate.