Study
Questions for Quiz #1
(Parenthetical info will not
appear on quiz – ya gotta
know to include those parts!)
|
Survey of Correlation
Models |
Briefly
describe the members of each of the major kinds of correlation/regression
questions and their interrelationships. (be sure to include the bases for
their inter-relationships) |
|
Briefly
describe the members of each of the major kinds of correlation/regression
questions and the significance tests used for each. |
|
3 |
|
|
Correlation & RH: Testing |
Be
prepared to describe the patterns of data that result in various combinations
of size and direction of r for both quantitative and binary X variables. |
|
Describe
the consequences of "weak" vs. "strong" nonlinear
relationships upon the meaningfulness of r values. |
|
Distinguish
between “comparing our results to our RH:” and “inferential accuracy of our
results” and tell the likely reasons for the different outcomes of each. |
|
Describe the things one
must be careful/specific about when comparing correlations. |
|
|
Bivariate Regression |
Describe
linear regression as a "linear transformation" and describe how
this transformation functions. How
does this function highlight that the data analyst must correctly decide
"which regression model to use"? |
|
Distinguish
“univariate prediction” from “bivariate prediction” and tell when to use
each. Describe the advantage of a significant
binary predictor over univariate prediction. |
|
|
Multiple Regression &
Prediction |
Describe
the statistical analyses that should precede your multiple regression
analyses and what you would be looking for when doing each. |
|
Carefully
distinguish between the proper interpretations of regression weights for
simple and multiple regression models. Describe how to interpret the multiple
regression weights of continuous vs. binary predictors. |
|
What is “under
specification” and what difficulties might it cause when using multiple
regression? What are the solutions to
these problems? |
|
Describe “proxy variables”
and tell what difficulties they might cause. What is the relationship between
“under specification” and “proxy variables”? What are the solutions to these
difficulties? |
|
Describe
the three attributes of “the multiple regression model” for a criterion variable
and explain why the third attribute is critical. What usually prevents the
discovery of “the model” and what should we do for a particular analysis and
for our programmatic research? |
|
What
is the risk when you decide not to further consider a predictor in your
multivariate analyses? Tell the things you should consider before making such
a decision (opdef,
simple, full mreg, various mreg,
curvilinear, interactions, indirect). |
|
|
Some Details About Multiple Regression |
Tell the various things
that influence the calculated values and significance tests of simple
correlations and simple and multiple regression weights. Discuss the use of standardized weights to
evaluate the relative contribution of the predictors to the model. |
|
Why do we conduct both
bivariate and multivariate analyses of the relationships between a set of
predictors and a criterion? What are
the possible results of this “dual analysis”. |
|
What are suppressor
variables (remember to distinguish the two kinds), how are they identified,
interpreted and “dealt with” in the present and future studies? |
|
Describe collinearity, tell
how to distinguish among different levels of it, and tell the problems
related to each level. What are the
ways of dealing with the problems caused? |
|
What is "range
restriction" and for what reasons does it happen? What problems does it cause for multiple
regression analyses? What are the
suggested solutions to it, and what problems might they cause? |
|
Briefly describe the
“regressions to remember” (multivariate power ,null flooding,
extreme collinearity, collinearity patterns).
What is the source of the surprise in each. If we get surprised, which should we
believe, the bivariate or the multivariate result and why? |
|
|
Statistical Control |
Distinguish
between experimental (be sure to mention randomization, matching, holding
constant, and balancing and how they relate to each other) and statistical
control and tell the correlation, regression and ANOVA models used to apply
the latter. What are the advantages and disadvantages of statistical control? |
|
Describe
the use of "residualization" to apply
statistical control. Include a specific example with the necessary formulas
to describe and carry out the procedure for each of the six statistical
control models you know (partial, multiple partial, semi-partial, multiple
semi-partial, ANCOVA & nested models). |
|
|
Multiple Regression Research Hypothesis Testing |
Describe
the alternative ways of asking if a particular predictor has a relationship
with a criterion that is unique from a set of other predictors. |
|
Discriminate
full and reduced models and tell the various approaches to generating the
latter, commenting carefully upon the relative quality of each. |
|
Describe
substitution analyses and tell how they differ from the construction and
comparison of "nested models.”
Also, describe the different reasons for proposing a substitution, and
how these might relate to the intent of the analysis (descriptive vs.
predictive). |
|
Describe the theoretical
and applied reasons for comparing regression models across populations. Discriminate between the comparison of
“fit”, “substitution” and “structure” when comparing models in this way. |
|
Describe the theoretical
and applied reasons for comparing regression models across criterion
variables. |
|
Describe the known problems
of each of the forward, backward, and forward stepwise statistical modeling
procedures? What does it mean to say that the decisions at each step in these
procedures are both statistical and numeric and it is the numeric part that
causes the problems? |
|
|
Path Analysis |
What is path analysis? What is the “starting point” of a path
analysis and what kinds of research hypotheses can we
and can we not test? How can a knowledge of path analysis help when designing
research and analyzing regression models (temporal/causal, indirect &
suppressors)? |
|
Path analysis is also often
called “causal analysis”. Can path analysis be used to study causal
relationships among variables? What do we get out of a path analysis that we
don’t get out of a multiple regression using the same variables? |
|
How does a “mediation
analysis” differ from a “path analysis”? Of what must we be careful in any
path analysis, but especially so when doing a mediation analysis? |