Research Methods Flashcards
What is a fact?
A statement about a direct observation that is so consistently repeated, virtually no doubt exists as to its truth value.
What is a theory?
A collection of statements which together try to explain a set of observed phenomena
What is a hypothesis?
A clear but tentative explanation for observed phenomena which can be tested
What do theories do?
- define
- organise
- interrelate
- explain
- make predictions on which hypothesis can be based
What are the qualities of a hypothesis?
- testable (can a test be designed to adequately test it?)
- falsifiable (can a hypothesis be disproven)
- all terms clearly defined
- rational (must fit in with what we already know)
- parsimonious (favouring simple definitions over complex ones)
What is a construct?
- theoretical concepts formulated to serve as causal or descriptive explanations
- can’t measure concepts until they are turned into operational definitions
What is an operational definition?
a concrete version of a construct
Constructs vs variables
Constructs are defined by theoretical definitions
TD: intelligence is the capacity to acquire and apply knowledge
Variables are defined by the operational definition
OD: The score on the standard test of intelligence
Types of variables:
Normal
Ordinal
Interval
Ratio
Normal variable
No way to put the categories in order e.g. gender, star sign
Ordinal
Variables can be ranked/ordered
e.g. position in a race (not necessarily regular intervals)
Interval
Data goes up in discrete regular (equal) intervals but there is no real zero e.g shoe size
Ratio
There is a real zero and there are equal increments between each data point e.g. height/age
what is an experimental design?
A design which allows yo to make causal inferences about the influence of one variable on the other. Allows us to determine causal relationships.
What is the independent variable?
The variable you are changing
What is the dependent variable?
The variable that is being measured
Features of independent variables:
- can have different levels
- can be between or within subjects
- can be a mixture of between and within if enough levels (mixed design)
What is wrong with between and within designs?
between: subject to variation
within: subject to order effects
What must you do in a between subjects design?
- must randomise the participants in each group
- stops you biasing your experiment
- allows you to carry out powerful statistical test for analysis
What is the within experimental design also know as?
repeated measure
Within experimental design
- each subject acts as their own control
- fewer participants needed
- order effects tho….
- counterbalance to avoid this
Factoral design:
- investigation of two independent variables on one dependent
- shows us the effect of each IV independently and also the interaction of the two
e. g. -morn shift and booze - eve shift and booze
- morn shift and no booze
- eve shift and no booze
Name the two different types of factoral design:
- within (each participant exposed to all four levels)
- mixed –IV1 between –IV2 within
What is a quasi experiment?
An experiment where there cannot be random allocation of groups e.g. when testing gender
What is a true experimental design?
Where the experimenter has complete control and is able to allocate groups randomly
What is the problem with quasi experiments?
-it is difficult to control extraneous variables so difficult to infer causality
How to battle the problem with quasi experiments?
Match using the pair design
e. g. if want to test the prevalence in 18-25 yr olds in prison communities vs. general pop
- extraneous variables could be Q and level of education, so get a bigger non convict group and try to match the IQ and education level of each prison person to one normal person.
Sometimes in within design experiments you can’t counter balance. what should you do?
Split your groups in half and have a control group. it is now a mixed design experiment.
Explain the developmental experimental design:
When you want to measure a variable with age/over time
-between (cross sectional) -age group of 5 and age group of 10
Within (longitudinal)- measure the same kids when they’re 5 and again when they’re 10
What are the problems with the cross sectional design?
- can’t randomly assign
- solution = matching
What are the problems with the longitudinal design?
- can’t counterbalance
- solution: control group
What is the point of counterbalancing and random allocation?
-gives an even distribution of errors across IVs
Talk about extraneous variable s
They are undesirable variables that add considerable error to our experiments. we want to eliminate or at least control their influence.
What are confounding variables?
Variables that disproportionately effect one level of the IV more than the other
What is a type 1 error?
Finding a relationship between the IV and DV when there isn’t one
What is a type 2 error?
Not finding a relationship between the IV and DV when there is one
Confounding variables are bad because they are
threats to internal vailidity
What types of confounds are there?
- Selection
- History
- maturation
- Instrumentation
Describe the effect of selection:
- bias the results due to the non random selection and allocation of participants to each IV level
- results when subjects assigned to each IV are systematically different in a way that will effect the DV
- especially tricky in quasi designs
Describe the effect of history:
When an uncontrolled event takes place between the two testing occasions
Describe the effect of maturation:
- can come from growing up (did dyslexia improve because of training or just age) or from having done the same test twice
- intrinsic changes in the participant between testing occasions
Describe the effect of instrumentation:
The change of measuring instruments accuracy or reliability over the course of an experiment
What is reactivity?
- because they know their being experimented on/watched the participant may act in a different way than they normally would
- can be the participant trying to please the experimenter (demand characteristics)
- can be the experimenter inducing bias from their expectancies
What can reactivity effect?
The internal validity!! if it influences on level of IV more than other
How do we reduce reactivity?
blind experiments - experimenter or participant or both (double blind)
What stems from reactivity?
- experimenter bias
- demand characteristics
Name the types of measurement errors:
- random (obscure results)
- constant (bias results)
What are the two criteria necessary to satisfy causation?
- sufficiency
- necessity
What is sufficiency?
Y is adequate to cause X
if you only have one variable you are showing necessity, not sufficiency
What is necessity?
X must be present to cause Y
What is precision?
Consistency –> reliability
What is accuracy?
truthfulness/correctness –> validity
Name the forms of reliability:
- inter rater
- parallel (if we test the same thing in a different way will our results match up?)
- test reliability (if we administer the test on two separate occasion will we get the same results - used for things we expect to stay the same e.g. personality, not mood - beware of order effects!!)
What is reliability continuity?
internal consistency - are all elements of our test (e.g. questions) measuring the same construct
How do we test reliability continuity?
Internal consistency:
By ‘split test reliability’ where the test is split into two halves and given to P on different occasions. similar results = good internal relibility
Forms of internal validity:
- content
- face
- critereon (concurrent/predictive)
What is content validity?
Does the test measure the construct fully? (all aspects of it or just certain parts)
e.g does the RM exam cover all aspects of the course
What is face validity?
Does it look like a good test?
e.g. does the content of the RM exam reflect the knowledge expected of the students?
What is criterion vailidity?
When you compare your results with an experiment that has already been done. do they correlate?
- concurrent -do the results of the test correlate with the old one (e.g. correlation between scores on two different depression tests or exams in different subjects)
- Predictive- does the test predict outcome on another variable (e.g. ambition and $$ in the future if you measured $$)
How do we measure construct validity?
- best measured over time but in the short term can measure using:
- convergent validity - should correlate with similar constructs
- discriminant validity - shouldn’t correlate with unrelated constructs
e. g. if you make a test to measure happiness results should correlate with contentment but not depression.
What is info about populations and samples called
parameters
statistics
Why do we sample?
time
access
money
sufficiency (pattern of results doesn’t change much even if we use entire pop)
What are the different types of sampling?
Random Systematic Stratified Snowball Opportunity/convenience
What is random sampling?
Each member of the pop has an equal chance of being selected
Usually quasi random
Not everyone picked will participate
What is systematic sampling?
Fixed intervals
e.g. if every 10th participant was female
What is stratified sampling?
Organised sample:
- Disproportional - specified groups are picked in equal proportions, not representative of population group size
- Proportional - specified groups appear in the same proportions as in the population
What is a cluster sample?
Researchers sample an entire group or cluster from the population of interest e.g. a school class
-can introduce bias
What is an opportunity sample?
People who are easily available are sampled
-can introduce bias
What is snowball sampling?
Recruit small number of Ps and then get them to recruit other people
-biases but useful if you want to recruit specific populations e.g. people with a speech impediment
What are the types of external validity?
-Ecological
(does the behaviour measured reflect naturally occurring behaviour?)
-Population
(is our sample representative, did everyone complete the study coz if not random selection becomes self selection by Ps)
What type of validity do naturalistic designs give?
High external
low internal
What type of validity do experimental designs give?
High internal
Low external
What does good ecological validity allow us to do?
Know if it will work/can be applied to real life
What does good population validity allow us to do?
generalise our results
