Using everything we’ve learned so far about the central limit theorem, the z-score, and hypothesis testing, we can now also perform proportion testing!
There are just a few new concepts to add into the mix:
The preliminary terrors – notation & terminology
p = the proportion of items that falls into H0
q = the proportion of items that don’t fall into H0
p̂ = the sample proportion
μ = μp̂ = p
σ = √ pq
σp̂ = σ /√ n
Z-score = p̂ – μp̂ / σp̂
Before we begin
You can only do this test successfully IF:
np̂ > 10 and nq̂ > 10
If this test fails you will first need to take more samples until both conditions hold true…
And a final note… this test can be used to evaluate against an actual population and its parameters OR a hypothesized population and its parameters.
A simple example
Your company has been having some problems in manufacturing plastic bags – they keep splitting. The equipment on the production line has been re-calibrated, and it’s expected that this will resolve the issue. You’ve been asked to validate (to a 95% confidence level) that the number of split bags has been reduced to below 4%. The quality control department took a random sample of 400 bags and 13 were found to be split.
In other words we cannot reject the null hypothesis that defects are >= 4%, because the probability (P = 21.78%) is considerably greater the desired confidence level (α = 5%).