SPSS 教程 回归分析.ppt

Marketing Research
Regression analysis
Last week we started to deal with correlation and an interesting application, factor analysis. This week we will deal with:
A little revision, then
The correlation between one dependent
variable and several independent (multiple
regression)
Today’s topics
When pare the data of two groups of people, or data from one group over two variables, we generally have two ways of making parison (other than univariate differences such as different standard deviation and so on). These two key descriptors are:
Differences or similarity in the absolute value of the
means, and
comparison of the movement of the scores between
two variables or two groups of the same variable
Bivariate analysis
T-tests, ANOVA
Correlation
Dependence techniques are used when some
variable(s) depend on some other variable(s).
This will be the topic dealt with THIS week
Interdependence techniques are used to help to
reduce many variables to a more manageable
number, and to understand underlying
meanings, through grouping similar variables
together. We have done this last week
Types of multivariate analysis
Types of multivariate analysis
Interdependence
techniques
Are inputs metric?
Cluster analysis
Metric multi-dimensional scaling
Factor analysis
Yes
No
Multivariate analysis of variance
Conjoint analysis
Dependence
techniques
How many dependent?
Multiple dependents & independents
Canonical analysis
Several dependent
One dependent
Multiple discriminant analysis
Multiple regression
Non-metric
Non-metric
Metric
Metric
Multivariate analysis of variance
Conjoint analysis
Dependence
techniques
How many dependent?
Several dependent
One dependent
Multiple discriminant analysis
Multiple regression
Non-metric
Non-metric
Metric
Metric
The coefficient of correlation for two
variables, x and y, is shown as:
xy.
It is always between
–1 (a perfect, but negative relationship)
and +1, a perfect, positive relationshi