For the graph to represent your data most effectively, consider the following guidelines.
The sample size should be between 20 and 1000
A sample size that is less than 20 may not provide enough power to detect significant differences between your sample data and the normal distribution. A sample size that is greater than 1000 may provide too much power. When a test has too much power, small and possibly meaningless differences between your sample data and the theoretical distribution appear to be significant.
The sample data should be selected randomly
In statistics, random samples are used to make generalizations, or inferences, about a population. If your data were not collected randomly, your results may not represent the population.
