Download - Review from Last Week
Review from Last Week
Appropriate for all types of research, all 4 types of Scientific Method
For any area of research Political Science, Physics, Economics…
Basics of Research design Anthropology to Zoology
Conducting Scientific Research
The Goal is Inference: Generalizability
The procedures are public Replicable
The conclusions are uncertain “Statistics is never having to say you’re certain.”
Follow the rules of inference We’ll learn these as we go
Components of Research DesignThe Basic Steps
A) The Research QuestionB) The TheoryC) The ModelD) The DataE) The Use of the Data
A theory includes HypothesesHypothesis: A Statement of What we
believe to be factual.
Independent Variable (X1)
Dependent Variable (Y)
Independent Variable (X2)
Y=f(XX11,XX22))
Good Hypothesis should:
Have explanatory power State Expected Relationship & Direction if
Possible Be Testable Written as simply as possible Relate to general, not specific
phenomenon Be plausible
Z is ANTECEDENT
Z X Y
Z is INTERVENING
X Z Y
SPURIOUS RELATIONSHIPS
X
Y
?
We hypothesize that X leads to Y, but the true relationship is that another factor is causing both.The only way we see this is by reasoning in our model and in our theory. Just looking at the data, we cannot uncover the causal relationships at work.
Alternative Hypotheses and Null Hypotheses
Two are compliments, not strictly opposites. HA and H0 are: Mutually Exclusive & Exhaustive HA: X is true
H0 : X is not true. HA: X is related to Y
H0 : X is not related to Y HA: X is positively related to Y
H0 : X is negatively related or not related to Y.
Example: Average score on the stats exam is 70. Our class has an average of 78. We can test the hypothesis that our class average was higher just because of sampling error and the hypothesis that our class average was higher because we have smarter students
A hypothesis is a statement about a relationship between variables. The null hypothesis H0 states there is no true difference between scores in the population. The alternative hypothesis Ha, is that the difference in our sample is truly reflecting a real difference in the population, that the difference is not due to sampling error.
All hypothesis testing is done against the null hypothesis
The Null Hypothesis H0 is the result you could get by chance.
The Alternative Hypothesis Ha is your research hypothesis. It is what you believe will happen.
Positive and Negative Relationships
Positive As X increases Y
increases Or As X decreases Y
decreases Two go in the same
direction
Negative (or inverse) As X increases, Y
decreases Or As X decreases, Y
increases
The Model A basic summary of our theory, specifying the
relationships among all the relevant factors Answers the research question by explaining the
Dependent Variable Is a representation of real world Outlines the hypotheses we believe and will try to
test DIAGRAM on the next slides should clarify the
relationships.
Example - Question, d.v., level, i.v.s, hypotheses
Each circle is a variable: Independent variables pointing to the dependent variable
Each arrow is a hypothesis about the relationship between variables (causality)
Overall, model represents part (or all) of our theory
Level of Analysis
(we implicitly make these decision when we chose the Dependent variable)
Choose: Level of Analysis Choose: Unit of Analysis Choose: Cases How do we do this?
Begin by asking: What is our population?
Building a Model II, Getting to Data
Cases will all be at the same levelBill, Susan, George, Henry...
81st Congress, 82nd Congress, 83rd….
Canada, France, USA….
Bill, Susan, Suffolk County, Cuba, Bill last year…
Getting to Data…
• What will your population be?• Your sample of cases should be
representative of the population.• When thinking about your cases be
obsessively specific!• What will qualify as a case?• What is the time frame?
Concepts
Part of our theories Define as clearly and concretely as
possible Link to Empirical phenomenon
Makes much easier to defend.
Variables Empirically observable characteristics of
some phenomenon Varies across cases 3 ways to discuss a Variable:
Where it fits in the model Whether or not it is observed How it is measured.
1. Where it fits in the model•Independent
•Dependent
•Intervening
•Antecedent
2. Is it observed?• Latent
• Manifest.
3. How it is measured OPERATIONALIZATION
convert abstract theoretical notions into concrete terms, thereby allowing measurement.
OR… process of applying measuring instrument in order to
assign values to some characteristic or property of the phenomenon being studied.
OR… TURN CONCEPTS INTO VARABLES and then into
DATA
Rules for Variables More possible values is usually better Mutually Exclusive - a case can hold only
one value You can’t be both tall and short
Exhaustive - Every Case has a value If a case changes over time so that it
holds different values of a variable… you should?
Measurement
Creating variables often requires creativityApproximate concept that you wish to
measure.How to measure abstract concepts?
- also depends on level of analysis.
Types of Operationalization Non-orderable Discrete Categories
A.k.a. Nominal Categories, names E.g., gender
Orderable Discrete Ordered, but not precisely ordered E.g., professor quality
Dummy, Dichotomous, 0/1 “Qualitative variable” Could fall into either of the above Presence or absence of something
Interval Consensus on differences between the units E.g., temperature
Ratio Scale Same as interval but with an absolute 0 point
Example of Levels of Example of Levels of MeasurementMeasurement
Suppose you wanted to measure smoking.• Ordinal: How often do you smoke?
Never 2-3 per day 1 pack per day > 1 pack per day
• Interval: How many cigarettes do you smoke each day?
• (What’s the level of analysis here? How would you define smoking for other levels of analysis?)
http://www.douglas.bc.ca/psychd/handouts/measurement_scales.htm
DATAChoose cases based on level
Represent population we want to generalize about
Collect facts about each of our variables for each of our cases.
V 1 V 2 … V K
Case 1Case 2 …
Case n
Cases
Are
Rows
Variables are columns
Examples of Measurements
www.freedomhouse.org/research/freeworld/2000/table1.htm
www.transparency.org/documents/cpi/2001/cpi2001.html