Review 2
When to use each test?
T-test (one sample)
- Quantity of variables: 1
- Types of variables: continuous
- Null hypothesis: some value (usually sample mean) = population mean
T-test (independent sample)
- Quantity of variables: 2
- Types of variables: 1 categorical (with 2 categories), 1 continuous
- Null hypothesis: population mean for category 1 = population mean for category 2
T-test (paired sample)
- Quantity of variables: 2
- Types of variables: 2 continuous (same variables but different time periods)
- Null hypothesis: population mean at time 1 = population mean at time 2
Note: Observations should be paired across time.
One-way ANOVA
- Quantity of variables: 2
- Types of variables: 1 categorical (with 2 or more categories), 1 continuous.
- Null hypothesis: population mean for category 1 = population mean for category 2 = population mean for category 3 = …
Two-way ANOVA
- Quantity of variables: 3
- Types of variables: 2 categorical (with 2 or more categories each), 1 continuous.
- Null hypothesis: No interaction between categorical variable 1 and categorical variable 2
Note: In other words, our null is that one categorical variable does not influence the effect of the other categorical variable on our continuous variable.
Pearson’s correlation
- Quantity of variables: 2
- Types of variables: 2 continuous.
- Null hypothesis: Pearson’s correlation coeff. for the population = 0. Variable1 and variable2 are not correlated.
Simple linear regression
- Quantity of variables: 2
- Types of variables: 2 continuous (usually)
- Null hypothesis: regression coef. for the population = 0. In other words, our independent variable does not predict our dependent variable.
Note: The key point here is that you have a dependent and an independent variable, meaning that you assume the direction of the relationship. Therefore, regression is more appropriate for predictions.
Chi-square test for independence
- Quantity of variables: 2 (or more)
- Types of variables: 2 categorical
- Null hypothesis: Variable1 is independent of Variable2.
Practice questions
Open the “school_data.sav” data set.
Choose the appropriate test for the following questions:
- Did schools improve their average scores between 1999 and 2000?
- Does average class size influences average achievement in the year 2000?
- Is there a relationship between percentage of parents who have completed HS and achievement in a given school?
- Is there a relationship between percentage of free meals in a given school and whether the school has summer classes?
- Do California schools with summer classes (year round schools) have higher average achievement?
- Is the mean achievement for the schools in the sample a good representative of the mean achievement for all California elementary schools?
- Suppose percentage of free meals are associated with achievement. Does the existence of summer courses in a school influences this relationship?
- Is average student achievement different between schools with different percentage of free meals?
Hint: Look carefully at the variable labels in your data. Note if the variable is continuous or categorical.