README
¶
Hypothesis Testing
While working as a data scientist, you will encounter the need to test many hypotheses. For example, you may want to know if a change to a website produced more email registrations, or you may want to know if a new model is actually performing better than an old model. These hypotheses must must be testable for you to actually come to some conclusion. In other words, there must be some way of determining if your hypothesis explains an observation.
Notes
- A "null hypothesis," usually denoted by H0, is the hypothesis that observations result purely from chance, the opposite of the alternative hypothesis.
- An "alternative hypothesis," usually denoted by H1, is the hypothesis that observations caused by some non-random process.
- A "p-value" measures the probability of observing a test statistic as extreme as the current test statistic.
- Hypothesis tests are carried out assuming the null hypothesis is true.
Links
Hypothesis Testing - Penn State
The four steps of hypothesis testing
Code Review
Chi-Squared Table
Calculate Expected Frequencies
Calculate Chi-squared Test Statistic
Output the Result of the Test
Exercises
Exercise 1
Design a test to determine if die is fair. Use the Chi-squared test statistic and a 5% level of significance. Answer the following:
- What would you measure?
- How would you calculate the test statistic?
- What range of Chi-squared values would cause you to abandon your null hypothesis?
All material is licensed under the Apache License Version 2.0, January 2004.
Directories
¶
| Path | Synopsis |
|---|---|
|
Sample program to calculate expected values.
|
Sample program to calculate expected values. |
|
Sample program to calculate a chi-squared value.
|
Sample program to calculate a chi-squared value. |
|
Sample program to output the result of the test, based on a critical value.
|
Sample program to output the result of the test, based on a critical value. |