Term
| Least Squares (OLS)- regression |
|
Definition
| sum of squares of the residuals |
|
|
Term
|
Definition
r - calculated from a random sample of two variables.
measures goodness of fit about the linear least-squares regression line for the observed values of the dependent variable in the sample |
|
|
Term
| F-test (usage) in regression |
|
Definition
| tests the statistical significance of the entire regression |
|
|
Term
|
Definition
| Unexplained variation of the residuals |
|
|
Term
|
Definition
|
|
Term
| Differences between bivariate and multivariate regression (2-variable and general linear model) |
|
Definition
Adding 3rd or more variable
assumption of no perfect multicollinearity |
|
|
Term
| Residual Term in regression (ESS or BSS or SST) |
|
Definition
ESS or BSS - difference btw mean Y and predicted Y - explained.
(treatment variation, explained sum of squares, variation between samples, the variation due to Factor A) |
|
|
Term
|
Definition
difference between Y and predicted Y - unexplained but accounted for
(random variation, residual sum of squares, the variation within samples, the error variation) |
|
|
Term
|
Definition
| BSS+WSS - sum of squares of the residuals with respect to the mean |
|
|
Term
|
Definition
| unexplained and unaccounted for either because of the lack of inclusion of a variable (violation) or inadequacy of the sample |
|
|
Term
| Why use variances in ANOVA |
|
Definition
| Allows the test of the null hypothesis that of the relationship between the datasets when sets are not the same type: eg ordinal to interval, categorical to ratio, ordinal to ratio, categorical to interval |
|
|
Term
| Concept of the standard error of the regression |
|
Definition
| The average error in predicting Y |
|
|
Term
|
Definition
H0: mu1=mu2...mux (x=number of categories - BSS)
H0: mu1=mu2...mux (x=number of rows - WSS) |
|
|
Term
|
Definition
1. Populations normally distributed
2. Samples are independent
3. The variances of the populations must be equal
4. The groups must have the same sample size |
|
|
Term
| General Linear (multivariate) regression assumptions |
|
Definition
1. Non-autocorrelation (non-serial correlation or non-autoregression)
2. Homoscedascticity
3. Zero means
4. Nonstochastic
5. Correct specifications
6. Stochastic error term
7. No perfect multicollinearity |
|
|
Term
|
Definition
| Error termes do not have a relationship with each other. Knowing one does not allow prediction of another |
|
|
Term
|
Definition
| variance is constant and random effects are throughout the data set. Var (ui) = E(ui2) = σ2 |
|
|
Term
|
Definition
| normal distribution: error terms normally distributed around the regression line?? Will cancel each other out [E(ui) = 0] |
|
|
Term
|
Definition
| error terms are not related to independent variables |
|
|
Term
|
Definition
| Y = B1 + B2Xi + ui or Y-hat = B1 + B2Xi – correct variables and formula, include all relevant variables (no violating assumptions or leaving relevant variables out) |
|
|
Term
| no perfect multicollinearity |
|
Definition
| independent variables are independent |
|
|
Term
|
Definition
| tests differences between two or more population means. An extension of a t-test |
|
|
Term
|
Definition
Best Linear Unbiased Estimate
1. Accuracy (Efficiency) - the variance of the sampling distribution of beta-hats will be the minimum for any estmators of beta
2. Beta-hat fits on beta |
|
|
