Unit 2: Ch. 9 Flashcards
causality
many quantitative research questions are about cause and effect
criteria for making causal inferences? (3)
- the cause must precede the effect
- there must be a demonstrated association between the cause and effect
- the relationship between the presumed cause and the effect cannot be explained by a 3rd variable or confounder (“intervening variable”)
- there can’t be an outside thing influencing the other 2 things
know what’s a cause and what’s an effect
cause comes before the effect (ex: smoking causes lung cancer)
what are the 3 types of quantitative research?
- experiment
- quasi-experiment
- non-experiment (observational studies in medicine)
experiment
offer the strongest evidence on whether a cause (intervention) results in an effect (outcome)
-this is why experiments are high on the evidence hierarchy (strong, but not as strong as systematic reviews)
3 characteristics of a true study?
- intervention
- researcher does something to one group but not the other - control group
- has no intervention but may have a placebo - randomization
- researcher assigns subjects to groups randomly
- “random assignment/selection”
- randomization is the “great equalizer”
quasi-experimental
involves an intervention but lacks one of the other 2 characteristics (either RANDOMIZATION or CONTROL)
non-experimental
a study w/ no intervention
-considered weaker studies than experiments or quasi-experiments; still valuable b/c some variables can’t be manipulated (ex: abuse)
examples of non-experimental studies (9)?
- non-correlational design
- prospective correlational designs
- retrospective studies
- time related study
- cross sectional design: data collected at only one point in time
- longitudinal study: data collected over time - correlation is an association between 2 variables; one may not cause another but one may happen w/ another (correlation ≠ causation)
- prospective design may also be called cohort study by medical researchers (go forward in time)
- retrospective correlational design is when an outcome is present, such as depression
- comparison studies (2 types): one is between subjects (ex: men and women in a study - compare men to women) and a within subjects design (men and women in a study; only compare women w/in the group)
- descriptive: simply describing what’s there
Purpose of control and randomization is to make the groups equal in regards to all factors except the?
intervention
quasi-experiment involves an intervention but?
lacks one of the other 2 characteristics (either randomization or control)
counterfactual
what would happen to people if they were exposed to a causal influence and were simultaneously not exposed to it
most causes are not ____; they only increase the likelihood that an effect will occur
deterministic
John Stuart Mill developed 3 criteria for establishing causal relationships. What are those 3 criteria?
- temporal: a cause must precede an effect in time
- relationship: must be an association between the presumed cause and the effect (ex: an association between smoking and lung cancer)
- confounders: relationship can’t be explained as being caused by a 3rd variable
what are the 3 techniques of research control?
- context: conditions of the study; want the conditions of the study to be held constant (when data is collected, where data is collected, and how data is collected)
- need control over the study context
- communication w/ participants is standardized, using scripts as necessary
- intervention must be delivered according to a plan “treatment fidelity”)
- have to have control over the context (who, what, when, where, and why of your study) - participants: typically done through randomization; also done through homogeneity (want participants to be as much alike on characteristics as possible)
- can be done through matching, statistics and ANCOVA (way of controlling characteristics statistically) - confounders: intervening/extraneous variables; confounding variables not part of the study but can influence findings
- researcher must be aware of what the confounding variables are/might be and control them
- ex: studying stress - there are other things that result from stress that look like the cause
biologic plausibility
evidence from basic physiologic studies that a causal pathway is credible
T/F: except for description questions, questions that call for a quantitative approach usually concern causal relationships
true
other features of quantitative research designs (3)?
- masking/blinding: when participants aren’t told anything about the study; done to eliminate/decrease bias
- there is a debrief and review after - time frames: researcher has to decide when and how often data will be collected
- location: decisions have to be made about where the data is collected
* quantitative studies consider these every time!!
pretest-posttest design
involves the observation of the outcome (mood) at 2 points in time: before and after the intervention
involves collecting pretest data on the outcome before the intervention and posttest data after it
control group
refers to a group of participants whose performance on an outcome variable is used to evaluate the performance of the experimental group on the same outcome
a control group is used for comparative purposes represents a proxy for the ideal counterfactual
experimental group
the group getting the intervention
randomization
every participant has an equal chance of being included in the group
aka random assignment
randomly assigned groups are to be expected to be comparable, on average, with respect to an infinite number of biologic, psychological, and social traits at the outset of the study
group differences on outcomes observed after randomization can therefore be inferred as being caused by the treatment
posttest-only design
the most basic experimental design
involves randomizing people to different groups and then measuring the outcome
baseline data
pretest data
posttest data
outcome data
placebo
pseudointervention
presumed to have no therapeutic value, which is also called an attention control condition (the control group gets attention but not the intervention’s active ingredients)
carryover effects
when subjects are exposed to 2 different treatments, they may be influenced in the second condition by their experience in the first
protocols
stipulate exactly what the treatment is for those in the experimental group
delayed treatment
i.e., control group members are wait-listed and exposed to the experimental treatment at a later point
validity
accuracy. Whether or not the questionnaire is measuring what it reports to measure
the degree to which inferences made in a study are accurate; in instrumentation, it is when an instrument measures what it purports to measure; accuracy of measurement
Hawthorne effect
A term derived from experiments at the Hawthorne plant, in which various environmental conditions (e.g. light, working hours) were varied to determine their effects on worker productivity
regardless of what change was made, productivity increased
thus, knowledge of being in a study may cause people to change their behavior, thereby obscuring the effect of the research variables
time series design
involves collecting data over an extended time period, and introducing the treatment during that period
correlation
an interrelationship or association between 2 variables
can be detected through statistical analyses
T/F: correlation doesn’t prove causation
true
cohort design (prospective design)
observational studies w/ a cohort design start w/ a presumed cause and then go forward to the presumed effect
retrospective correlational study
an effect (outcome) observed in the present is linked to a potential cause occurring in the past
case-control-design
cases w/ a certain condition, such as lung cancer, are compared to controls w/o it
descriptive correlational
researchers seek to describe relationships among variables, w/o attempting to infer causal connections
self-selection
selection bias
attrition
loss of participants over time
homogeneity
only people who are similar w/ respect to confounding variables are included in the study
confounding variables in this care are not allowed to vary
matching
involves consciously forming comparable groups
statistical conclusion validity
the ability to detect true relationships
internal validity
extent to which the IV, tx, or cause influenced the DV, outcome, or had an effect
“Internal validity = how much the Independent variable is Influencing the outcome”
external validity
generalizability (whether nor not you can apply the findings to other settings/people)
construct validity
degree to which key concepts are captured in the study; whether or not everything should be measured are; are constructs accurate?
stastical power
refers to the capacity to detect true relationships
can be achieved in various ways, the most straightforward of which is to use a large enough sample
threats to statistical conclusion validity (3)?
- low statistical power (e.g. sample size too small)
- power analysis - weakly defined “cause” - independent variable not powerful; tx not intense enough to have an effect on the outcome
- unreliable implementation of a tx - low intervention fidelity
power analysis
researchers can estimate how large their samples should be for testing their research hypotheses through power analyses
threats to internal validity (5)?
- temporal ambiguity: too much or too little time between tx and outcome
- selection threat: bias from preexisting differences
- ex: studying postpartum moms - including first time moms and moms w/ 2 or 3 kids (selection bias b/c first time moms and moms w/ previous kids are different) - history threat: co-occurring events that influence outcome
- maturation threat: developmental changes that influence outcome
- comes up a lot in pediatric and adolescent studies (gero studies too) - take age into consideration
- ex: outcome in a 4yo may be totally different than in an 8 or 10yo - mortality/attrition threat: loss of participants so the group make up changes
threats to external validity (1)?
inadequate sampling of study participants: may be the wrong type of people
-ex: clinical sample and non-clinical sample; must compare apples to apples so that it’s generalizable
threats to construct validity (3)?
- is the intervention (IV, tx) a good representation of the underlying construct?
- is the intervention or awareness of the intervention that resulted in benefits?
- does the dependent variable (DV, outcome, effect) really measure the intended constructs?
