Term
Response variable: Categorical (2 groups)
Explanatory variable: None |
|
Definition
z-test (one-sample)
testing a proportion
assumptions - randomized collection of data, and proportion of success > 15 and proportion of failure > 15. |
|
|
Term
Response variable: Quantitative
Explanatory variable: None |
|
Definition
t-test (one-sample)
testing a mean
assumptions - randomized collection of data, approximately normal population distribution (only if n < 30). |
|
|
Term
Response variable: Categorical (2 groups)
Explanatory variable: Categorical (2 groups) |
|
Definition
z-test
testing two proportions
assumptions - randomized collection of data, for z-test: proportion of success and failure in group 1 each are > 5, and proportion of success and failure in group 2 each are > 5. For CI: proportion of success and failure in group 1 each are > 10, and proportion of success and failure in group 2 each are > 10. For McNemar's test (dependent samples): b+c>30 (failure before and failure after). |
|
|
Term
Response variable: Quantitative
Explanatory variable: Categorical (2 groups) |
|
Definition
t-test
testing two means
assumptions - independent, randomized collection of data, approximately normal distribution (if either first og second group has n < 30). |
|
|
Term
Response variable: Categorical (>2 groups)
Explanatory variable: Categorical (>2 groups) |
|
Definition
Chi-Squared test
testing for dependence
assumptions - randomized collection of data, expected count is >5 in all cells. |
|
|
Term
Response variable: Quantitative
Explanatory variable: Categorical (>2 groups) |
|
Definition
ANOVA (analysis of variance)
testing several means
assumptions - normal distributions within each group, same standard deviation across all groups (works reasonably, if the SD is the same within a factor of two), independent samples, randomized collection of data. |
|
|
Term
Response variable: Quantitative
Explanatory variable: Quantitative |
|
Definition
Linear regression
assumptions - mean of y depends linearly on x, randomized collection of data, for each x the population values of y should follow a normal distribution with the same standard deviation for all x.
Model checking: calculate standardized residuals (outliers?), histogram of standardized residuals (normal distribution), normal quantile plot of standardized residuals (normal distribution), scatterplot of residuals vs. x (linearity and constant standard deviation). |
|
|
Term
Response variable: Quantitative
Explanatory variable: Categorical and/or quantitative |
|
Definition
Multiple regression
assumptions - all y are normally distributed with the same standard deviation σ.
µy = α + β1 x1 + β2x2 + · + βkxk.
Calculate standardized residuals (outliers?)
Histogram of standardized residuals (normal distribution)
Normal quantile plot of standardized residuals (normal distribution)
Scatterplot of residuals vs. each x (linearity and constant standard deviation)
Scatterplot of residuals vs. predicted values (linearity and constant standard deviation) |
|
|
