Research Midterm Flashcards
Logical steps of scientific method
Step 1) state problem or pose question
Step 2) form a hypothesis (explanation or possible answer)
Step 3) Test hypothesis
Step 4) Analyze data
Step 5) Draw Conclusions
Independent variable
what the researcher wants to study
Dependent variable
what the researcher will measure to provide evidence
extraneous variables
factors other than the independent variables that can affect the dependent variables
Confounding variables
extraneous variables that cannot be controlled for or eliminated from an experiment
Hypothesis vs theory
hypothesis
- prediction of the final outcome of research
- concrete, specific statement
- based on deductive reasoning
Theory:
- belief or assumption about how things relate to each other
- establishes a cause and effect relationship between variables with a purpose of explaining and predicting phenomena
- based on inductive reasoning
Inductive reasoning
- observe: collect info based on one or more examples
- Analyze: ID patterns
- Infer: create a general rule from observations
- Confirm: make more observations to strength inference
inductive reasoning can be used to:
- part to whole: the whole is assumed to be the same as the individual parts (ex. all 10,00 dogs examined had fleas, therefore all dogs have fleas)
- Extrapolations: areas of study beyond the exact area studied will behave in the same manner (ex. all dogs have fleas, cats are similar to dogs, therefore they must all have fleas)
- Predictions: future will behave the same as the past
Deductive reasoning
- starts with a rule or law and then deduces specific examples from that law
- using scientific method, a law can be determined for a specific case and then further testing is done, to show that it holds true in other circumstances
Research Hypothesis
- statement indicating the researchers true expectations as to the outcome of the study, may be the same as or opposite of the null hypothesis
- expectation as to outcome of the study
- based on understanding of research problem and analysis of what is known from related literature
- follows directly after the problem statement
- never proven, more supported or verified
Statistical hypothesis:
- Null hypothesis: statement indicating expectation of no relationship or no difference for purposes of statistical testing
- alternate hypothesis: a statement indicating expectation of the researcher that one group differs from another as a function of the independent variable
Compare groups
- treatment groups receives intervention
- control group receives placebo
Null: groups do not differ on dependent measure
- treatment and control group score the same on balance test after 10 weeks training
Alternate: groups differ on dependent measure
- treatment group scores significantly higher than control group - directional alternate hypothesis
- treatment group scores are significantly different than control group - non-directional alternate hypothesis
reject vs do not reject null
- Reject the null means that real difference exist between the groups (if a = .05 and p = .049, reject the null)
- Do not reject the null means that no group differences exist ( if a = .05 and p = .52, do no reject the null)
Clinical relevance vs effect size
- often discussed when results of a study are non-significant
- researchers attempt to make a case for the effect size being large enough that it is likely to be of interest in clinical practice
- with no objective test for clinical relevance, researchers must make the case with logic
- can be important when results are statistically significant but the effect size is small
types of research
basic, applied, translational
Basic research:
- conducted purely for discovery of new knowledge, with little regard for whether there is an immediate application for that new knowledge
- pure, fundamental research
- theoretical in nature
- takes many years for the results to find practical utility
Applied research:
- driven by need to find a solution to a specific problem
- improved products or processes
- infers beyond group or situation studied
- interpretation of results relies upon basic research
Translational research:
- specific to studies in which a finding from basic is first investigated in humans
- common for research related to clinical applications to start with animal subjects to ensure safety
Quantitative vs qualitative
Quantitative research:
- dependent variable are numbers
- tradition or positivist approach
- clearly stated questions, rational hypothesis
- developed research procedures
Qualitative research:
- dependent variable are quality (descriptions)
- typically anthropological and sociological research methods
- observations of a natural setting
- in-depth descriptions of situations
- interpretive and descriptive
Experimental vs non-experimental
Experimental:
- investigator provides treatment or intervention
- confirms or refutes cause-effect relationships
- aims at achieving understanding
- involves measuring effects of independent variable on dependent variables to determine a mechanism
Non-experimental:
- no treatment or intervention
- cannot speak to cause and effect
- aims to provide a detailed, useful description of characteristics of a population, group, or individual
- is often hypothesis generating
Cause-effect relationship
- strong correlation between the cause and effect
- the cause precedes the effect
- cause always produces the effect
- there is not a viable alternative explanation for the effect
- independent variable is the cause and dependent variable is the effect
Experimental research designs
1) Single-subject research (not a case study)
- use when there is a high degree of individual variability in performance
- single or small group of participants
- unlike case study, researchers provide an experimental intervention or treatment
Experimental research designs
2) Pre-experimental designs
- weak designs
- no random assignment of participants to groups
- little control of threats to validity
- sometimes used due to practical constraints
Experimental research designs
3) True experimental designs
- require random assignment of participants to groups
- control many threats to internal validity, including history, testing, stats regression, selection bias
- referred to as randomized control trials or RCT
Experimental research designs
4) Quasi-experimental designs
- designed to optimize external validity
- used when circumstances prevent random assignment of participants to groups
non-experimental/descriptive methods
- patterns and trends in relationships among individual characteristics and specific disease or health conditions
- data acquired through survey or existing data bases (no intervention)
- large # of participants used
- can answer questions like “does obesity status lead to higher rates of cancer?”
Analytic vs descriptive
analytic descriptive research:
- no intervention
- describes patterns and trends, but quantifies the relationship or association between variables
- quantitative research
Non-analytic descriptive research:
- no intervention
- describes patterns and trends without numerical data
- qualitative research
Non-experimental research designs
1) Cohort study (analytic)
- Prospective: how does current exposure lead to outcome disease in the the future? (looks forward)
- select diverse groups from population who vary in exposure to risk factor
- follow groups over substantial period of time, periodically checking for outcome
- no direct intervention
- quantifies the probability of developing disease based on exposure to determinants: relative risk
Non-experimental research designs
2) Case control study (analytic)
- Retrospective: given current disease or health state, what was exposure in the past? (looks backwards)
- create groups with (case) and without (control) disease state
- survey groups on past exposure to risk factors
- quantifies odds of developing health condition based on inclusion in case of control and exposure to risk factor: odds ratio
Non-experimental research designs
3) Cross-sectional study (analytic)
- observes current occurrence of disease or condition in a given population
- “snapshot”
- quantifies disease prevalence
Non-experimental research designs
4) Epidemiological research (analytic/non-analytic)
- study of determinants of incidence, distribution, spread, and control of disease and health
- “why are some people healthy and other are not?”
- can be similar to both case control or cohort studies but is often more generalized
Non-experimental research designs
5) Case study (non-analytic)
- single-subject report
- often seen in medical journals for unique & interesting condition or first occurrences of a treatment
Non-experimental research designs
6) correlational research (analytic/non-analytic)
- form of non-experimental or descriptive research
- conducted to evaluate statistical relationship between two or more variables
- sometimes conducted for purposes of prediction or hypothesis generation
empirical vs theoretical research
empirical:
- collects data
- uses observations/measurements to create new ideas
- experimental
- follows scientific method
theoretical:
- uses knowledge and logic to create hypothetical idea
- ideas are not tested in lab
- exmaples include math modeling and proofs
primary vs secondary research
primary:
- collection of new data or creation of new idea
secondary:
- uses previously collected data to make generalizations and/or to strengthen current understanding of topic
- descriptive reviews, systematic reviews, meta-analyses
validity definition
ability of research study or instrument to faithfully reflect true state of variables being studied in the population of interest
internal vs external validity
Internal validity:
- ability to conclude that only the independent variables affected any differences in measures of the dependent variables across groups or across test on the same group
- is a measure of the success of efficacy of the effect of the experiment
External validity:
- ability to apply results of a study to the sample population in a real world setting
Threats to internal validity
- *History**: something occurring over time that could affect the dependent variables
- *Maturation**: effects related to the passage of time, such as aging
- *Testing**: beneficial practice effects for repetitions of the same test
- *Instrumentation**: negative effects of reliability problems with equipment or observers
- *Statistical regression**: tendency of extreme scores to regress toward the mean upon retest
- *Selection bias**: comparison groups are not equal at the beginning of the study
- *Experimental mortality**: loss of participants from a study (for any reason)
- *Selection-maturation interaction**: maturation affects groups within the study differently
- *Expectancy**: rater expectations influence data
Threats to external validity
Interaction effect of testing: pretest changes the group’s response to experimental treatment
Interaction of selection bias and experimental treatment: biased sample produces skewed results not representative of population
Reaction effects of experimental setting: some element of setting causes modification of participant behavior
Multiple treatment interference: experiencing one treatment affects participant response to a subsequent treatment
Ecological validity
- extent to which testing conditions in a study are like conditions in the environment being studied
- another component of external validity
Types of Reliability
Test/Retest: use the same test on the same group of subjects more than once and compare
Parallel Forms: compare the equivalence of two tests that measure the same thing
Internal consistency: measure of how well each individual item of the test represents the whole test
Sampling
- who we include in study
- how they are chosen
Importance of sampling
- goal is to collect data from participants in research study and to be able to generalize the findings to the larger population that the participants represent
- Used to answer two types of questions:
- 1) is there a relationship between variables?
- 2) is there a difference across conditions?
Sample size
- large enough to be representative of population
- small enough to enable practical collection of data from the standpoint of time & resources
- often calculated to arrive at a power level of 0.8, where power is the probability of correctly rejecting the null hypothesis when it is false
sampling techniques
1) random selection: all members of the population have equal chance of being selected
2) stratified random: all members of designated subgroups have equal chance
3) stratified proportional: equal proportion of each subgroup is randomly selected
4) systematic: every nth individual on a list is selected
5) convenience: qualified readily available selected
6) cluster: population is separated into groups then a sample of clusters is selected
Sampling criteria
inclusion criteria:
- characteristics individuals must possess to qualify for participation in a research study
Exclusion criteria:
- characteristics individuals must NOT posses to qualify for participation in a research study
Ideal study design
- sample is representative of pop of interest
- groups of samplers are as close to identical as possible at the start of the experiment except for the treatment being applied