Hypothesis Testing is a statistical tool, that allows the resemblance of two or more process attributes – mean, median and standard deviation. It provides with a method to determine differences.

  • Important part of a conclusion reached on the basis of random sampling (statistical inference)
    • A hypothesis is a statement that we want to verify using data
      • is there are difference?
      • has there been a change?
  • Null and Alternative hypotheses are formed
    • Level of risk and confidence required
  • Experimentation is conducted and interpret the results
    • Fail to reject the null hypothesis
      • No change. No difference, or…
    • Reject the null hypothesis
      • There is a change, a difference

What might we “Test” for?

We will be testing for a Change or Difference in Process…

  • Central Tendency
    • Mean, Median, Mode
  • Variation
    • Variance, Standard Deviation
  • Proportion
    • % (ratio, proportion)
  • Frequency (of occurrence)
    • Distribution of count/frequency

Testing Protocol

  • Null” hypothesis (H0)
    • This is a statement validating the status quo – there will be no change of significance observed. Any differences detected are purely due to chance and not a change in the process.
    • Symbols:
      • = (equals)
      • < (not less than)
      • > (not greater than)
  • Alternative” hypothesis (H1)
    • This is a statement that there will be a difference of statistical significance detected, there has been a change.
    • Symbols:
      • <> (doesn’t equal)
      • < (is less than)
      • > (is greater than)

The hypotheses are complementary to each other. If one is true, other is not true, and vice versa.

If the p is low, the null must go!

When performing any statistical test, outcome is based on sampling from population, therefore there is a room for an error. Most statistical tests are run with a 95% confidence level, indicating that there is a 5% chance of making an error.

The decision of whether to simply accept or reject the null hypothesis is predicated on the calculated p value. If the p value is a smaller amount than or adequate to a preassigned significance level (normally set at 5%), then we reject the null hypothesis and accept the alternative.

A p value are going to be calculated by the statistical software when running a hypothesis test.

Types of Hypothesis Testing

There are many different types of hypothesis tests, they can be divided into two main categories:

  • Parametric tests
    • Makes inferences about parameters like mean and variance
    • Based on assumptions of specific distributions (ex. “normal” or “t” distributions)
  • Non-parametric tests
    • Makes inferences about frequency distribution like median, distribution type
    • Usually include sign and rank tests (type of “math” used)
    • Do not require assumptions of normality (but do have some assumptions… always check them!)

Depending on the type of data you have collected, as well if it’s normal or non-normal, there are several hypothesis tests available for comparing process characteristics.

Hypothesis test decision tree

Leave a Reply