sampling at Achmad Mardiansyah's Journal

Archive for the ‘sampling’ tag

understanding: confidence interval, confidence level and number of sample

Just want to share about an idea of understanding confidence interval, confidence level, and number of sample without being bothered by the formulas. I found many people are still confused about this.

here it is…

Suppose you join a shooting tournament, and your goal is to shoot a circle target in a given distance.

Shooting analogy	Statistics in research
Goal: to shoot a circle target in a given distance	Goal: estimate mean of population in the form of an interval. Why interval? Because you don’t do a census. Therefore, the best thing that you can do is providing an interval, and hoping that the mean of population is in the interval
jury will mark your shoot	audience will comment on your research
Your gun: best sniper rifle	Your tool: doing sampling
Assume: environment is ideal: wind, rifle, bullets	Assume: sampling process is ideal
Size of target: measured by diameter	Mean of population: defined by interval. This is the confidence interval
you’ll feel more confident if the target is wider.	in here, the higher the confidence level, the wider the interval will be. Because if you provide a wider interval, the higher probability of population’s mean appear in the interval. That’s why you feel more confident
unfortunately, the jury won’t allow you to widen the target size	the audience won’t like a wider estimation interval. They’ll think that your research is not accurate.
so, the target size is fixed, and so your confidence level.	here, you have to stick with your confidence level (let say 95%). You don’t want to be ashamed right?. You need to find out, how to shorten the estimation interval without decreasing confidence level.
one way to make sure that you can hit the target is by shooting more	here, you can produce a better result by increasing the number of sample because the more data the closer to the reality. Then, your audience will trust you more.