Why random samples are preferred to nonrandom samples?
Answer:
It eliminates any source of human bias, that is a person doesn't pick which results matter, or pick what indivdual readings demonstrate a hypothesis - purely because they do.
This ensures the results are more reliable.
Random samples provide for a better representation of the whole, while nonrandom samples can be too specific to represent an entire quantity.
Because of selection bias, we tend to select samples (either consciously or unconsciously) that we think will give us the answer we either want or expect. Random sampling avoids this problem.
I have 3 popsicles, cherry, orange and lime. Which melts fastest? So I take one of each out of the fridge and weigh them. After 20 minutes, I weigh them again. These are my results: Cherry 93%, Orange 91% and Lime 88% (of their initial weight). I repeat the experiment 50 times.
The final averages are Cherry 92.5% Orange 92% Lime 91.5%. So would it matter to you if I had always measured the Cherry first and the Lime last? Or do you think I should have randomized the order in which I weighed them? Non random sampling introduces a bias in the experiment. In my example it introduced a longer time (possibly - I didn't really say exactly) between the start of the experiment and the measurement. The bias is another variable which has to be included in the degrees of freedom for the experiment. For small sample sizes random may NOT be preferred to a non-random rotation. For larger sample sizes, random is always preferable so that the "laws" of statistics can be applied - most of which assume random distributions. Random actually means pseudo-random but that is another subject.
Random samples are preferred because they provide a better sample of the population.
Your samples are chosen randomly, therefore not in any way manipulated... which can skew your results.
In the scientific method, you are trying to demonstrate that there is a pattern to how nature works. The pattern, algorithm, or hypothesis relates to only a small part of nature, not to everything.
The collection of data that represents All of nature tends to be a random distribution. So, if you pick out only the data that represents only your hypothesis, then you are artificially stacking the deck.
Instead, you want to pick a representative sample of nature, a random sample, and apply your statistical filter to it to show one algorithm to represent underlying patterns.
The answers post by the user, for information only, FunQA.com does not guarantee the right.
More Questions and Answers: