Measuring Behaviour Etc Flashcards
Operational definition
- the phenomena to be measured should be defined in terms of the operations used to measure them
- a good operational definition is reliable and reproducible
- loss of contract validity can occur (conventional meaning may be lost)
- eg: alcoholism, PMS, ADHD etc
- essential in experimental research
Ostensively definitions
-phenomena to be observed should be carefully described (textually, graphically, photographically) and examples can be given
-common in observational research
(Basis of ethograms and systematic behaviour)
Nominal scales
- categories, taxonomies, typologies
- eg: male/female; liberal/conservative/democrat; ADD+H/ADD-H; extroverted/introverted
- types of statements: x is different from y; assignment of labels
- purely qualitative information
Ordinal scales
- continuum or spectrum of observations.
- ranking within a category is possible; often different names and certainly different quantities
- absolute values not known
- eg: low, moderate, high self-esteem; ranking of a race
- types of statements: x is greater than y; assignment of values
- qualitative with some crude degree of quantity (ranked)
Interval scales
- has no absolute zero point, i.e. arbitrary
- eg: Celsius scale (0 is a relative point on the scale, when ice melts; it is a convenient convention). Likert scales
- common in psychology and neuro: rating scales, from 0-10 (where 0 does not mean no liking at all), IQ scale (0 does not mean no intelligence)
- quantitative; we know by how much the values differ
- spacing between values is known, but a score 120 is not necessarily double a score of 60
Ratio scales
- has an absolute zero point (absence of the quantity can be indicated)
- zero means zero
- eg: Kelvin (0K is the absence of heat
- common in psych and neuro: from no response to n degree of response
- eg: score on memory test
- quantitative; we know how much of the quantity exists
Scales of measurement: the implications
- nominal and ordinal can be analyzed with a number of non-parametric statistical analyses
- interval and ration analyzed with parametric stats
Environmental/situational variables
- variables referring to the manipulation of the environment
- eg: treatment, tasks
Instructional variable
- variables referring to what is told to a participant, instructed, or simply suggested
- applies mostly to human research and animal research
Subject, participant, or individual variables
- quasi independent variables
- differences in individuals
- eg: ethnicity, habits, diseases etc
- important in idiographic vs nomothetic research
- not true independent variable because it is not manipulated
Controlled variables
-any variable that is controlled or held constant across all treatment conditions of experiment
Extraneous variables
- “threatening variable”
- “obscuring factors”
- impact on dependent variable that is not ind variable
- variables that are not being studied
Confounding variables
- an extraneous variable (usually unmonitored) that can inadvertently affect another experimental variable
- potential source of variability (experimenter-expectancy effects
Solutions for confounding/extraneous variables
- random assignment of participants to treatment conditions
- keeping extraneous variable constant
- matching participants on extraneous variable
- building the extraneous variable into study or blocking so that the extraneous variable becomes an independent variable
- statistical control of variable: ANCOVA so that extraneous variable become a covariate
Experimental research designs and variables
goal: causal relationships
Variables: IV=manipulated by experimenter; DV=measured by experimenter
Issues: internal validity (confounding variables); external validity: generalizability, ecological validity.
Correlational research designs and their variables
Goal: correlational relationships
Variables: predictor=used to make a prediction; criterion=predicted variables
Issues: directionality problems A->B, B->A, AB?
Third variable problem: not AxB, but AxC
Threats to interference
- experimenter-expectancy effects; confirmation bias
- demand characteristics: subjects own expectancies; can occur in animals also
- subject predisposition effects
- cooperative-subject: want to provide data that will support hypothesis
- screw you effect: want to sabotage experiment
- evaluation apprehension: want positive evaluation; hypothesis independent
- faithful subjects: do what they are asked
- placebo effect
Control groups
- control groups: do not get the independent variable (manipulation, treatment, etc)
- placebo: control group that receives dummer treatment (sugar pill) that is assumed to have a positive effect
- nocebo: opposite of placebo… effect of IV is worse not better
- sham group: surgery with no experimental purpose
- quasi control: subjects in this group dont get placebo
- sub type: simulator groups - subjects are asked to act as if they had received treatment. Useful in double-blind experiment to detect and assess experimenter-expectancy effects and demand characteristics
Single-blind procedure
Minimized demand characteristics; not always possible because of informed consent requirements
Double-blind procedure
-reduce demand characteristics and experimenter-expectancy effects
Partial-blind procedures
- experimenter not in the know until just before the treatment is about to be administered
- can minimize experimenter-expectancy effects
Debriefing
-asking subjects about what they think happened to identify problems with demand characteristics
Basic techniques for experimental control
- single blind
- double blind
- partial-blind
- deception
- disguised-experiment technique/unobtrusive experimentation
- debriefing
Multiple researchers control
- having multiple experimenters, observers, handlers, trainers etc
- useful if you suspect experimenter effects
Experimenter-expectancy control groups
- 3 groups of researchers formed
- group 1: led to expect one experimental outcome
- group 2: led to expect opposite experimental outcome
- group 3: led to expect no effect
Unrelated-experiment technique control
- separate the presentation of the IV over the DV. TWo experiments are necessary (within design so subjects do both)
- experiment 1: subjects receive IV
- experiment 2: subjects receive DV
- subjects are told the two experiments are not connected
Experiment with all control groups
1: receive 1 mg of melatonin
2: receive 3 mg of melatonin
3: control - placebo (sugar pill)
4: quasi control: no placebo
Random sampling
- single random sampling
- Stratified random: random sampling within segments (strata) of the population based on specific characteristics are considered (age, gender etc)
- cluster sampling: naturally occurring units of individuals (groups) are randomly selected. Eg. Children in day care group, students in this class.
- proportionate sampling to avoid over representation
Non-random or non-probabilistic sampling
- haphazard sampling (not true random sampling)- example with grabbing fish from tank
- convenience sampling
- volunteer sampling
- systematic sampling (every nth element sampled)
- sequential sampling: gradual, one at a time
- quota sampling: stratified without randomness (based on convenience sampling principles)
- purposive/selective: selection based on pre-determined criterion
Selective sampling, non-sampling, and strong criterion based sampling
- sometimes look for exceptional individuals, highly qualified
- eg: police dogs, traffic controllers, med students
Random sampling
-selecting subjects from a population so all members have same chance of being selected for a sample (and all possible sampled have the same chance of being selected)
Random assignment
- using random chance to determine which condition a participant will experience
- determines difference between experimental and quasi experimental research
