What is it about?
We provide a non-technical explanation of how Type I error can be split in to tails. We provide two examples from industry, one marketing example, the temperature of coffee served. Second example is provided from manufacturing industry, in the form of a length of a bolt. The bolt can be too long or too short. We have an error with too long of a bolt or we can have a problem with to short size of a bolt. If the bolt is either too long or too short. In either case, the manufacturer experiences a different problem. In the paper, this problem and costs of rejecting a truer null hypothesis or the bolt being too long or too short are explained. Likewise, we also explain serving the coffee to hot or too cold.
Featured Image
Why is it important?
In the paper, we explain what type I error is, how it can be split unequally. More importantly, we provide explanations of the cost of type I error or cost of rejecting a true null hypothesis, or cost of a too cold coffee or too hot coffee. In addition, in the manufacturing example we provide an example with bolt being too long vs. bolt being too short.
Perspectives
This is important to managers estimating the cost of type I error. In classical inferential statistics the type I error is always divided equally. For example in a typical problem probability of type I error or sometimes refereed to as alpha is .05 split between the right and left tail equally. In normal basic inferential statistics or hypothesis testing .025 on the left tail and .025 on the right tail. What we are proposing in this paper is to divide type I error unequally, say 04 on the left tail and .01 on the right tail still adding to .05. instead .025 + .025 = .05 it is .04 +.01 = .05. The cost of too hot coffee is higher than the cost of too cold coffee
Dr Ceyhun Ozgur
Valparaiso University
Read the Original
This page is a summary of: Unequal Division of Type I Risk in Statistical Inferences, Decision Sciences Journal of Innovative Education, January 2004, Wiley,
DOI: 10.1111/j.0011-7315.2004.00018.x.
You can read the full text:
Contributors
The following have contributed to this page
