Term
| Definitions of Research Design (2) |
|
Definition
Broad - The program that guides the investigator in the process of collecting, analyzing, and interpreting observations.
Narrow - A logical model of proof that allows the researcher to draw inferences concerning causal relations among the variables under investigation. |
|
|
Term
| Purpose of Research Design |
|
Definition
| Answer questions objectively, accurately, and in a timely manner. |
|
|
Term
| Natural Phenomenon vary in... |
|
Definition
Whether or not they exist
Which of the multiple states is present
Or to what degree they exist in a given situation |
|
|
Term
|
Definition
| Sum of the squared deviations about the mean of a set of scores |
|
|
Term
|
Definition
| Average squared deviation about the mean |
|
|
Term
|
Definition
| Variance in measures due to some known or unknown influences that "cause" the scores to lean in one direction more than the other. (There are real differences between the subjects.) |
|
|
Term
|
Definition
| Random fluctuations in measures due to chance. Not a real difference between subjects |
|
|
Term
| Total (observed) variance = |
|
Definition
| Systematic Variance + Error Variance |
|
|
Term
| Some systematic variance is labeled... |
|
Definition
| Error variance because of our ignorance of the causes of Y |
|
|
Term
| 3 Criteria for Causal Inference |
|
Definition
1. Temporal Order or Temporal Precedence
2. Co-variation of Cause and Effect
3. Internal Validity (No alt. explanation, non-spuriousness) |
|
|
Term
| Temporal Order/Temporal Precedence |
|
Definition
Must show that cause X occurs before effect Y.
Active - can always estab. TO
Attribute - can sometiems estab TO, other times can infer it, and sometimes you can't. |
|
|
Term
| Co-variation of cause and effect |
|
Definition
| to what extent to values of x co-occur with values of y, OR to what extent does the presence of one variable correspond to the presence of another. (Do not truncate (shorten too much)) |
|
|
Term
|
Definition
| We must show there is no alternative explanation for the observed co-variation between x and y. |
|
|
Term
|
Definition
| Model hypothesizing X1 having an indirect and not direct effect on Y |
|
|
Term
|
Definition
| Hypothesizes that X1 is not the original cause for Y, but is a conduit for its true cause X2. |
|
|
Term
|
Definition
| X2 occurs first, causing both X1 and Y |
|
|
Term
|
Definition
| There are other causes that cause Y (X2, X3, X4). You must show that X1 had an additional effect beyond them. |
|
|
Term
|
Definition
| Says you are partially right. Relationship only occurs under certain conditions |
|
|
Term
|
Definition
| Ability to generalize from the empirical study to the real world. |
|
|
Term
| Qualities of Good Research design |
|
Definition
1. Designed to answer research question
2. Maximize External Validity
3. Minimize Error Variance
4. Maximize Systematic/True Variance - allow variance to be observed.
IF CAUSAL HYPO, then these also
5. Establish temporal order variables
6. Control Systematic Extraneous Variables. |
|
|
Term
| 3 components of classic research design (used to meet 3 criteria of cause and effect) |
|
Definition
Manipulation
Comparison
Control |
|
|
Term
| Lesli Kish's 4 categories of variables for control purposes |
|
Definition
1. Experimental Variables - those in the study
2. Controlled extraneous variables
3. uncontrolled extraneous variables which are empirically related to our experiment(the ones we are really worried about)
4. uncontrolled extraneous variables which are treated as randomized errors. |
|
|
Term
|
Definition
Randomization
Limiting sample subset on control variable
Matching
Controlling in the statistical Analysis portion of the project |
|
|
Term
|
Definition
| the more variables you remove from group 3 (uncontrolled extraneous variables relating to experimental variables) the more confident you can be in your causal relationship inference. |
|
|
Term
|
Definition
random selection of subjects and treatment groups.
moves all from group 3. |
|
|
Term
|
Definition
| limit sample to a single subgroup, limiting variance, therefore that group can no longer affect your variance. cost is not being able to generalize to population |
|
|
Term
|
Definition
pair subjects based on the control variables and then randomly assign them to different groups.
Limited by finding matching subjects. |
|
|
Term
| Control in statistical Analysis |
|
Definition
| Put variable you want to control for in experiment as another independent variable, therefore controlling for it. Most often used in PLS. |
|
|
Term
|
Definition
argument
observation measurement analysis |
|
|
Term
| 2 types of research designs |
|
Definition
experimental designs
Ex Post Facto designs |
|
|
Term
|
Definition
power to manipulate the independent variable
power to randomly assign subjects and treatment groups
Strong in Causal inferrence (manipulatable) but weak in internal validity (too much control) |
|
|
Term
|
Definition
lack either power to manipulate or power of randomization.
weak in causal inference (temporal sequence unnkown) but strong in internal validity (Measurements made in real world, have chance to notice other relationships.
Most common in social sciences |
|
|
Term
| Extrinsic threats to internal validity |
|
Definition
| differential (nonrandom) recruitment of subjects. |
|
|
Term
| Intrinsic threats to internal validity |
|
Definition
Changes in subjects, measuring instrument, or reactive affect of observation.
History-all events that occurred during the time of the study that may affect it.
Maturation - developement - have a control group, both will develop the same.
Experimental mortality - drops out of study after pre-test but before post test.
Instrumentation - changes in measuring tool used
Testing - the process of being tested may change results.
Regression artifact - scores move closer to mean during post test. |
|
|
Term
|
Definition
| Multiple Independent variables 4 group design (first group getting X1 and X2, second and third getting just X1, and last getting nothing. |
|
|
Term
|
Definition
| 4 group experiment design solves the problem of X1 only being exist or not exist, can do magnitudes of X |
|
|
Term
|
Definition
| Multiple people same point in time (survey) |
|
|
