3.1 Hypothesis Tests Flashcards
What is a hypothesis
A testable idea
What is the difference between a null and alternative hypothesis?
The null hypothesis always states that there is no relationship between variables. The alternative hypothesis states that there is a relationship between variables, and often describes it.
Why do we reject H0 instead of confirming H1?
Because we can never be certain about the nature of the population. When we detect a change between two samples we can calculate how likely that is to happen randomly, but we don’t know enough about the alternative to be certain about why something changed.
What is the difference between statistical significance and practical significance?
Statistical significance is calculated based on observations and probabilities.
Practical significance can only be established with process knowledge.
What are type 1 and type 2 errors?
Type 1 Error (alpha risk)
Producer’s Risk
Reject a true H0
False positive
Convicting an innocent
Calling a good part bad
Type 2 Error (beta risk)
Consumer’s Risk
Fail to reject a false H0
False negative
Acquit a criminal
Call a bad part good
What is correlation and is it useful for determining root cause?
Correlation is the mathematical relationship between two variables. A strong correlation can suggest that they are linked in some way, but causation can only be established with process knowledge.
What is regression analysis?
Regression is a fitted line model on two correlated variables. It describes how strongly the data follows the equation of the line and therefore, how good the model is at predicting outcomes.
Which R values show the strongest correlation?
an R value of 1 is a perfect correlation, while 0 is no correlation.
What do we need to know about our data when choosing a hypothesis test?
How many samples we are comparing
Is the data attribute or variable
What type of statistic we are comparing (mean, std. dev, % of total)
