Chi-Squared Tests Flashcards
How do you calculate the expected value for a cell?
(Row Total x Column Total) / Sample Size
How do you calculate the chi-squared values for each cell?
(fo - fe) 2 / fe
What is the test statistic for chi-squared testing?
The sum of the chi-squared values
What is the formula for the number of degrees of freedom?
(m - 1)(n - 1) where ‘m’ is rows and ‘n’ is columns
What is the critical value?
Using the formula booklet table, find the value given for the number of degrees of freedom and given significance level
What is the conclusion if the test statistic > critical value?
Reject H0 and suggests there is association
What is the conclusion if the test statistic < critical value?
Reject H1 and suggests there is no association
What does a greater chi-squared values signify?
A greater chi-squared signifies a larger disparity between observed and expected values
Why can causation not be assumed if there is some association?
Causation cannot be assumed as a third variable may be involved
What does an expected frequency less than 5 suggest?
If they are less than 5, they are likely to be inaccurate because small samples create more random fluctuations
What is the solution to an expected frequency less than 5?
You combine categories suitably to give expected values greater than 5
What must you remember once you have combined categories?
You must re-evaluate the observed values as the categories for them have been combined too
What does chi-squared distribution assume about the data?
What does this mean for expected values less than 5?
Chi-squared distribution assumes that the sum of the squares are of normal distribution
If fe > 5, distribution of fo is asymmetrical so needs to be combined
What does a goodness of fit test, test?
Goodness of fit test tests whether a given set of data fits a particular distribution
What variables have an impact on the procedure of the goodness of fit test?
What is the impact?
When the mean is estimated for poisson and probability for binomial, you subtract one from the number of degrees of freedom
What is the recurrence relation for Poisson Distribution?
P(X = r + 1) = P(X = r) x (y / r + 1)
When do you combine categories?
When expected frequency is less than 5