Midterm 1 - Lecture 3 (CH4) Flashcards
How do we measure?
Operational Definitions
Operational Definitions
Operational Definitions: a concrete way to measure an abstract object
- We are defining the variables that we hope to study
In operationalizing, we’re defining variables we hope to study…
Variable
Variable: any event, situation, behavior, or characteristic that varies (ie. Isn’t constant)
* Need an operational definition to study something - YOUR PLAN FOR MANIPULATING/MEASURING
3 types of Variables:
- Participant
- Situational
- Response
Participant Variable (Types of Variables)
- Characteristics that individuals bring with them
- Can be measured, not manipulated
Situational Variable (Types of Variables)
Characteristics of the situation or environment
- Can be measured and/or manipulated (EX: observing vs. instructing)
Response Variable (Types of Variables)
2 components:
- Responses – can be measured only: performance on tasks (e.g. memory, math, perseverance, attention), reaction time, physiology (sweating, HR, BP, pupil dilation), self-report
- Behaviour – can be measured only: helping others, tipping in restaurants, donating money, smiling
What negative/impactful variable can also exist?
CONFOUNDING VARIABLES: Any variable that we are not interested in, that is intertwined with our variable of interest (its op def’n).
May impact the interpretation of our result
Types of Studies
Correlational Designs (Types of Studies)
- A type of non-experiment (no manipulation)
- Draw a sample, then measure 2 or more different variables: See how well they “hang together”, How co- related are the variables?, How strongly is one associated with the other?
(EX: Facebook use & lonelines)
Correlation
- A statistic that indexes the degree of relationship between two variables.
- +/- = “direction” of relationship: Positive (both increase), negative (one increases, other decreases), or non-existent (flat line)
- Number = “strength” of the relationship
- How closely is one variable associated with the other one?
Linear relationships can be summarized by a single
“number”:
- r (r is only useful for linear relationships)
- can range from -1 to +1
- -1 = perfect negative linear relationship
- +1 = perfect positive linear relationship
0 = no relationship
Correlation DOES NOT equal…
Causation; third variables only come out of correlations
Types of Studies
Experimental Designs
- Language: When can we say differences in variable A caused changes in variable B?
a. A must precede (happen before) B
b. When A is present, B must follow
c. When A is absent, B should not follow
d. When A is the only thing changed that might affect B
Internal Validity (Key Features of Experiments)
Internal Validity: the ability to infer that the IV causes changes in the DV
- Co-variation between two variables
- Temporal precedence
- Eliminate plausible alternative explanations
How do we work toward achieving internal validity?
(Key Features of Experiments)
- Experimental Control
- Random Assignment of people to condition
Experimental Control (Ways to Increase Internal Validity)
– Only the IV changes across conditions
– No “confounds” (a second kind)
- Confound (recap):
– A variable that is intertwined with the IV and may impact (co-varies with) the DV
– Could explain all or part of the result
Random Assignment (Ways to Increase Internal Validity)
A method of assigning subjects to experimental
and control group so that each subject in the experiment has the same chance of being assigned to any of the conditions
- Allows the researcher to balance out influence of all other (participant/situational) variables except the IV…
– Works “on average”
- More effective as number of participants increases
Random assignment to condition balances out misc. effects on the DV
Designing experiments that allow for casual inferences:
Experiments allow for casual inferences in part because they incorporate temporal precedence
- Also, attempting to eliminate the influence of all other variables except for the one that is being manipulated: Usually achieved by trying to ensure that every other feature of the control and experimental conditions is held constant, except for the IV
- In non-experimental, both variables measured at the same time - resulting in no info about what comes first
IMPORTANT.
3 requirements for casual inference (Experiments)
- Temporal precedence
- Need for covariation
- Eliminate plausible alternative explanations - which involve the possibility that some confounding variable could be responsible for the observed relationship
- Experiments do this through random assignment & different forms of experimental control
What’s a crucial part in achieving internal validity?
Keeping variables constant
Further criteria for claiming cause:
To be sufficient, the cause must always produce the effect
Non-Experimental Designs
Non-experimental designs: relationships are studied by measuring or observing the variables of interest
For any non-experimental method, both variables of interest are measured and none are manipulated
Allows us to detect covariation between variables, another term to denote some non-experiments is the correlational method: in which variables are observed but not manipulated, shouldn’t be confused with the correlation statistic
Interpreting the results of non-experimental designs:
Cannot directly tell us whether two things are causally related
- EX: discover that two variables are related, but not if one has a casual impact on the other
Also introduces the third-variable problem: possible additional variable for an observed correlation between two variables
Third Variable Problem:
Two variables aren’t directly related to each other
Since third variables offer an alternative explanation for the observed relationship, correlations alone can never be considered evidence of cause
