Navigational Menu OVERVIEW OF STATISTICAL THINKING |
MICROCASE BASIC STATISTICS REGRESSION REGRESSION--To
determine the (standardized or unstandardized) slope(s) of the
regression line which predicts a dependent variable given knowledge of
one or more independent variables.
For Asimple@
regression (i.e., one dependent and one independent variable), this
REGRESSION option reproduces the information available in the
SCATTERPLOT option without the actual scatterplot; REGRESSION also
provides standardized regression coefficients (the same as Pearson r
with two variables; the same as the partial correlation coefficient with
one dependent variable and multiple independent variables) as well as unstandardized regression coefficients
(expressed in raw score units). REGRESSION
is most useful when you have multiple independent variables, when you
desire a multiple regression equation or the multiple correlation
coefficient (R). 1.
When you select REGRESSION, the window screen will ask you for
your dependent variable and a list of independent variables (you must
supply at least one). As
with previous statistical options, you may elect to run the analysis on
a subset of the full sample. As
with the CORRELATION option, there is no way to eliminate outliers in
this version of regression analysis except through the subset function.
When you have indicated your dependent variable, your independent
variable(s), and any subset variables, click on OK in the upper right
hand corner of the window. 2.
The summary screen shows the BETA weight (standardized
multiple regression coefficient; partial correlation coefficient) for the relationship between the
dependent variable and each independent variable when the effects of the
other variables in the model have been eliminated or controlled.
The BETA can be compared to the zero-order Asimple@
correlation coefficient (r) which
describes the relationship between the dependent variable and the
independent variable when the effects of the other variables have been
allowed to impact the independent and dependent variables.
The summary screen also shows the R-squared which tells
you the amount of the variation in the dependent variable which can be
attributed to all the independent variables considered together. 3.
ANOVA (on the left side of the screen) provides the ANOVA summary
table used to test the probability that an R-squared of that magnitude
could occur by random sampling error alone.
ANOVA also provides a chart of the regression coefficients--both
standardized (BETAs) and unstandardized; this chart also provides the
standard errors and the t-values to allow you to test whether each of
the regression coefficients is statistically different from zero (i.e.,
is statistically significant). 4.
CORRELATION (on the left side of the screen) provides the
zero-order correlation matrix for all the variables in the regression
analysis. 5.
MEANS (on the left side of the screen) provides the number of
cases, the mean, and the standard deviation for each variable in the
analysis. for questions or comments contact me at mduncombe@coloradocollege.edu |