Variables/control Definition of IV
An IV a variable that is manipulated in an experiment
Variables/Control Definition of DV
A DV is a variable that is measured in an experiment
4 types of IV
Physical Experiences in the Environment for People
Physiological - manipulate a biological state
Experience - amount/type of training/learning
Stimulus/environment - aspect of environment manipulated
Subject/Participant - aspect treated like IV (gender, age)
4 types/measures of DV
The Correct Frequency is the Amount of Duration
Correctness - right/wrong
Rate/frequency - often
Degree/amount - (likert) how much
Latency/duration - how fast or how long something occurs
What is the difference between a NUISANCE and a CONFOUND variable?
A nuisance variable is UNWANTED, affects ALL. Often about the participant (history, gender, physical characteristics, etc.). RANDOM
A confound is UNINTENDED, SYSTEMATIC, BETWEEN groups and INVALIDATES experiment. BIAS
Best experimental controls (4)
- Randomize
- Elimination of specifical extraneous
- Constancy of extraneous across groups if you can’t remove
- Balancing of effect across all if can’t remove
Carryover effect
EVENT influences next responses
Order effect
Order POSITION affects responses
How do you completely counterbalance an experiment?
- Event is =# of times to participant
- Event is =# of times each session
- Event must precede and follow =# of times
What does every observed score consist of?
True score (hypothetical) + Error (random, bias)
Rosenthal effect
Experimenter expectancies: when expectancy influences participant scores.
single blind study
when experimenter doesn’t know who is control/who is iv group
pact of ignorance
Participant demand bias and good participant bias effect
What is response set?
In response bias, when context affects the way participant responds; can be setting or questions.
In response bias, can also be social desirability.
Sources of experimenter error and solutions
Random: noise, temp,time
Solutions for random: Standardize, balancing
Bias: (1) Experimenter characteristics
Solutions for 1: Standardize, balancing, replicate
Bias: (2) Experimenter expectations
Solutions for 2: Standardize, objective coding, single blind
Sources of Participant error and solutions
Random: careless, distracted
Solution: clear instruction, exp emphasis on accuracy
Bias (1) Demand: something about the study or questions cue the participant
Solution: Double Blind
Bias(2) Good participant effect: participant behaves like they think researcher wants
Solution: Deception
Bias (3) Response: yay/nay: always answer yes or no
Solution: random question order, include a reverse score question
Bias (4) Response set: bias of context
Solution: Pilot, review questions and setting
Sources of Observer error and solutions
Random: careless, distracted
Solution: use mechanical
Scorer bias: confirmatory - see what they want to see
Solution: mechanical, observable behaviors**, standard coding, make blind
What is construct validity?
Measure what you mean to measure to prove/disprove hypothesis
4 criteria for construct validity plus 1
- Reliability: test retest, inter rater/observer, internally consistent
- Content: questions capture all aspects of construct
- Convergent: do aspects correlate?
- Discriminant: things not the same should look different, do not measure high on opposing measures
- ALSO, sensitivity
How do I increase reliability, content V, convergent V, discriminant V?
Reliability: % that retest and interrater repeat, add more questions/check validity of questions
Content: have all dimensions and large enough set
Convergent: look at similar measures, known groups, other indicators
Discriminant: check for the disjointed answer patterns
How do I check sensitivity of measure?
Don’t restrict range, avoid all or nothing questions, add scale, do a pilot
Why do non experimental study?
Natural setting
Where manipulation not feasible
Establish association before experimenting
Rich data
How are experimental and non-experimental different?, why do non experiment
In non, there is no manipulation of IV
In non can compare size of association
In non, can predict/make selections (GRE)
In non, can study behaviour change over time
In exp, can prove cause and effect
In non, can only correlate.
5 types of DESCRIPTIVE studies
Archival
Case
Natural observation
Clinical observation
Participant observation
