|
Method |
Objective |
Description |
Dependent var |
Independent var(s) |
Comments |
| std multiple regression |
Predict values of Y given values of X |
Linear model Y =
b0+b1X1+b2X2 ... |
Interval |
Interval or binary |
Significance tests require normality of population
vars |
std
Logistic Regression |
predict the presence or absence of a
characteristic or outcome based on values of a set of predictor
variables. |
similar to a linear regression model but
i where the dependent variable is dichotomous
Models the log of the odds that Y=1 given value of X's
logit(p) = ln(p/(1-p)) = b0+b1X1+b2X2
the p's are unknown |
Binary |
Interval or binary Many stats
programs will all categorical indep var, which is internally converted
to dummies |
The effect of X1 on the odds that Y=1 is given by
exp(b1)=e^b1 |
| Multinomial logistic regression |
classify subjects based on values of a
set of predictor variables. This type of regression is similar to
logistic regression, but it is more general because the dependent
variable is not restricted to two categories |
f model
the odds that case i has Y=j rather than Y=k
log(πij/πik)
= b0 + b1X1 + b2X2 ...
|
Categorical |
Continuous or binary Many
stats programs will all categorical indep var, which is internally
converted to dummies |
Same goals and data requirements as discriminant
analysis, an older, less favored technique |
| Discriminant analysis |
|
Y is a categorical variable. We use
b0+b1X1+b2X2 to guess which category a case belongs to |
Categorical, including binary |
Interval or binary |
Uncool version of multinomial logistic regression |
| Poisson regression |
The dependent variable is freq of cases
in a cell of a crosstab (contingency table), and the explanatory
variables are factors and covariates. |
|
Categorical ("factors") and continuous
("covariates" |
|
|
