Lecture 2: Term 1 Principles of research design Flashcards
What is Research Design?
OS, unambiguous
Overall structure of the study to be carried out
Ensure evdience obtain from study allows to answer research Q as unambiguously as possible
Design is not i——— r——- to any method of d—– c——
intrinsically related
data collectio
A variable is something that can v—- or c——
may vary
N
iap
roem
Naturally
in a population
result of experimental manipulation
Validity
I
E
Ec
Internal (are my causal inferences valid)
External (how generalisably are my findings
Ecological (are my findings relevant to the real world rather than just for the experimental setting
Two types of categories of research designs
Experimental non experimental (no manipulation of variables)
Three feautures of experiments
EM
S
R
Experimental manipulation (set up 2+ experimental conditions that differ only with respect to the hypothesised cause of outcome of intrest)
Standardization - keeping things the same across conditions
Randomization
random allocation of participants to experimental conditions
unrelated and related designs a.k.a?
what is the third design?
why good?
why is this difficult?
independent groups (between participants)
Repeated measures(within participants) (counterbalance to control for order effects
Matched pairs eliminates order effects need many people to find matches (some data will be lost) very difficult to do well to focus on what you would match
Why is it important to distinguish between related and unrelated designs?
will inform which statistical test we choose
DIFFERENCE TESTS (not correlations)
related (same subjects in the two conditions)
- parametric = t tests for related samples
- non paramentric = wilcoxon matched pairs
unrelated (between participants)
- t test for unrelated samples
- Mann Whitney U test
correlations (pearsons product moment or np = spearman rho)
Types of experimental desgin (5)
RCT lab field true (invloe randomisation) quasi (like experiemnts by no strict ransomisation allocation to EC maybe due to social porcess or may not be ethical to maipualt IV)
Experimental design pros (1) cons (4)
possible to draw conslusion about causality
only deal with a small no of variables and answer simple questions
lab ex have low ecological validity
social by nature therefore experiementer effects
not always possible to manipulate variables
Non experimental designs
cross sectional
longitudinal
non-experimental evaluation
Cross sectional study def
pros and cons
Investigates a sample of ppts at a single point in time (answer descriptive questions, questions about relationships between variables, not well suited to testing causal hypotheses) do not control for unwanted influences like experiments but can bet addressed @ data analysis stage provided a good sample size_
larger scope = deal with large no of variables and east to add new variables to gather more info (vs ex designs)
useful for learning about the structer of psychologyical consepts
more naturalistic
non-response is a major methodolgical problem
ovvservations happen at the same time cross sectional studies are not well suited to answering questions about causality
Longitudinal study def
pros and cons
follows a group of ppts over a long period of time meaurements are taken at multiple time points
allows to study - change and factors influencing development
well suited to studiyng change
because assessments are made voer a period of time LS may addresse causal hypotheses however w/o experiemtnal control conclusions are tenative
2 major practical concerns - practise effects
drop-out
non experimental evaluation
an evaluation of a piece of research investigating the quaility and or effectiveness of an insitutions programme or service
set in the real world
allocation of peoplate to control groups may not be possible or affordable for an evlauation
Inferring causality
when a confounder completely explains the effect between X and Y it is said to be
a corelation between and y is not sufficient to infer a causal realtionship between the two
confounding variable may have causal effect on borh
spurious
