Midterm 2 Flashcards
What is a representative sample?
A sample that accurately represents the population that you are studying.
What is a biased sample?
A sample that does not represent the targeted population, contains different characteristics and the procedure favors the selection of certain individuals in comparison to others. Can occur by chance or by selection bias.
Example of biased samples?
Researcher going door to door
Financial compensation
Survey return rate
What do we use to calculate sample size?
Power analysis
What are the major factors that influence sample size? (4)
1) Size of the difference you need to detect
2) Variability in the factor of interest
3) P value you plan to use as a criterion for statistical significance
4) How confident do you want to be that you will detect statistical significance, assuming a difference exists
Effect size (power analysis)
The difference between mean (quantitative) or proportions
(qualitative) of two groups.
What is Standard deviation?
The variability within a sample.
What is a Type 1 error?
P level, usually 0.05 (5%).
What is power (power analysis)?
The probability of finding an effect (“how confident do you
want to be of your effect”), usually 80%.
What is the Direction of effect?
It can it go up only (one-tail); up and down (two-tail)
What are the types of Statistical tests? (3)
The chi-square, t-test, etc.
What is Attrition?
An adjustment in sample size (eg. death of animals, dropout)
What is the Resource equation?
E = (Total number of animals) − (Total number of groups)
What is the E variable? (resource equation)
The degree of freedom of analysis of variance (ANOVA)
When do you use the resource equation? (4)
When you cannot predict:
1) Effect size (ie. mean).
2) Standard deviation (ie. no previous findings available).
3) Multiple endpoints measured.
4) Complex statistical procedure used.
What are the two types of non-probability sampling?
1) Convenience sampling
2) Quota sampling
What is convenience sampling?
A sampling method where participants are easy/convenient to access/reach and are available to participate.
What is quota sampling?
Identify a subgroup, then choose from each subgroup to get a quota.
What is simple quota sampling?
Sampling method where little is known about the characteristics of a target population and equal number of participants selected for.
What is proportionate quota sampling?
Sampling method where demographics are known and proportionately represented in sample.
What is animal research quota sampling? And it’s pro?
Sampling method where buying animals from single supplier and thus animals obtained have same proportion on many characteristics.
Provides a sample that is representative of target population.
What are the four probability sampling methods?
1) Simple random
2) Systematic
3) Strata
4) Cluster
What is simple random sampling?
Sampling method where all individuals have an equal chance of being selected and each selection is independent of others.
Example: Choose 10 students from this class, give all students a number, then pick 10 numbers.
What is simple random sampling with replacement?
After each selection, he/she is added back into the pool and an individual can be chosen more than once.
What is simple random sampling without replacement?
After selection, the individual is removed from the pool and the probability changes with each selection. It does not produce independent selections and guarantees that no individual is chosen more than once.
What is the warning for simple random sampling?
When participant selection is left completely up to chance, can lead to a sample that is not representative of the population.
What is systematic sampling?
Sampling method where you choose every # participant on a list. Combination of simple random sampling + systematic process.
What is stratified sampling?
Sampling method where the population is divided into strata (subgroups), then participants selected from each strata.
What is simple stratified sampling?
Obtain equal numbers from each strata
* Equal representation
* Does not represent proportions in the population
What is proportional stratified sampling?
Select a different proportion (%) from each strata.
* Proportion depends on what representation in the actual population
* Difficult to compare between strata if some strata makes up a small percent of population.
Quota vs Simple stratified sampling? (similarities and differences)
-Similarity: Subgroups are grouped together
The difference is in random selection
-Quota sampling – not randomly selecting individuals from
population. Researcher chooses who gets sampled (individuals are pre-selected out of convenience)
-Stratified sampling – randomization first then selection
What is cluster sampling?
Sampling method where the population can be identified into clusters; then choose from a proportion of the clusters.
Advantage of cluster sampling? (2)
1) Quick and easy to get a large sample
2) Can measure all individuals in groups
The disadvantage of cluster sampling?
Some clusters are completely omitted, what if something is unique about certain clusters?
What research strategy do you use when examining individual variables?
Descriptive research strategy.
What is the descriptive research strategy?
Describes the current state of a variable(s) and does not examine the relationship between variables. (Not comparing between groups)
What research strategy do you use when examining the relationship of two variables?
Correlational research strategy.
What is the correlational research strategy?
Measure two (or more) variables for every individual/subject in the group and identify correlations between variables. Measures relationship between variables - As one variable changes, does the other variable also change? How does the variable change?
* Positive, negative, curvilinear
No manipulation of any variables
What do you use when examining the relationship by comparing scores for each group? (3)
1) Non-experimental research
2) Quasi-experimental
3) Experimental
What is non-experimental research? (3)
- One variable has two levels/conditions
- Lacks a control group
- No manipulation
What is experimental research?
*Establish cause-and-effect relationships
* Compare 2 or more groups on a dependent variable
* The group is manipulated
* Individuals are randomly assigned to the group
* Has a control group
What is quasi-experimental research?
*Similar to experimental
* Has a control group
* But groups cannot be randomly assigned
* In part because of preexisting participant/subject variable
* Does not provide a definitive answer about cause-effect
Difference quasi-experimental vs experimental
Quasi cannot be randomly assigned.
What are the two types of validity and their subgroup?
-Validity of measurement: face validity and construct validity.
-Validity of experiment: Internal validity and external validity
What is internal validity?
The extent to which the observed results represent the truth in the population we are studying.
What is external validity?
How well the outcome of a research study can be expected to apply to other settings.
Validity in nonexperimental research strategies?
- Nonexperimental research allows the researcher to examine the behavior in real-life settings
- Realistic environments offer higher external validity
- Environment is not manipulated or controlled ∴ limited internal validity
Validity in Experimental Research Strategies?
- Experimental research allows careful control of the environment
- Conducted in a controlled and constant lab setting ∴ high internal validity
- Experimental research often conducted in unfamiliar
environments, so harder to generalize beyond the experimental setting ∴ limits external validity
Threats to external validity? (3)
- Generalizing across participants or subjects
- Generalizing across study features
- Generalizing across different measures
What is generalizing across participants or subjects? (5)
- Selection bias
- Convenience sample
- Volunteer bias
- Participant characteristics
- Cross-species generalization
What is generalizing across study features? (3)
- Novelty effects
- Multiple treatment interference
- Experimenter characteristics
What is generalizing across different measures? (2)
- Sensitization
- Timing of measurement
Threats to internal validity? (2)
1) Obscuring variables
2) Confounding variables
What are confounding variables? (4)
An unmeasured variable that influences both the supposed cause and effect
* Assignment bias
* Environmental variables
* Time-related variables
* Observer bias
What are obscuring variables?
Variables that can cause null results or misses.
* Ineffective manipulation
* Measurement error
* Excessive variation in data
What are the criteria for generalization? (3)
- Controlled comparison rule:
* When comparing data between experimental and control groups, the only difference must be the independent variable - Sampling rule
* Adequate sample size to control for random sampling error and unbiased sample so it is representative of the population - Operational definitions rule
* Determines what principles can be generalized
* Single experiment only allows generalization within specific terms of the operational definition
What are exact replicas?
Repeat the experiment, and get the same results, and increase confidence in the findings.
What are conceptual replications?
The same construct, different experiments/methodology, get consistent results and conclusions get better generalized to entire constructs.
What are the four basic elements of an experiment?
- Manipulation
- Measurement
- Comparison
- Control
What are the two critical components of an experiment?
1) Independent vs Dependent variable
* More than one IV = factorial
* More than one DV = multivariate
2) Experimental vs Control group
What are the two approaches to comparing group differences?
1) Between-subject design: each group gets a different treatment.
2) Within-subject design: Subjects get both treatment
levels.
Advantages of between-subject design? (2)
1) Each measured score is independent.
2) Can be applied to wide range of research questions.
Disadvantages of between-subject design? (2)
1) Need more individuals.
2) Individual differences can lead to differences between groups even before any manipulation has occurred
What is variability?
Variability refers to the distribution of scores.
Variability of between-subject design?
Differences between treatments. We want large variability.
How to mitigate variance? (3)
Standardize procedures and treatment setting
Sample size (larger)
Group equivalency
What are the types of between-subjects tests? (2)
Posttest-pretest
Matched pair
What is the goal of posttest pretest?
The goal is to compare the treatment vs control group (not pre-test vs post-test)
What are the pros and cons of posttest pretest?
Pros:
* Test equivalence of groups
* Limit threats to internal validity
Cons:
* Time-consuming
* Sensitize participants to procedure
What is the matched pair design?
Each condition uses different participants but they are matched in terms of important characteristics.
What are the pros and cons of the matched pair design?
Pros:
* Minimize the effect of individual differences
Cons:
*Attrition affects both groups
* If B1 drops out, B2 must be taken out also
* Difficult to match people exactly
Within-subject design advantages? (2)
- Fewer participants are needed
- Eliminate variability due to individual differences
Within-subject design disadvantages?
- Participant attrition
- Cannot apply this design to all studies
- Time-related effects (history (environmental), maturation, instrumentation, order, sequence)
What is counterbalancing?
With counterbalancing, the participant sample is divided in half, with one half completing the two conditions in one order and the other half completing the conditions in the reverse order.
The goal of counterbalancing?
Use every possible order of treatments with an equal
number of individuals participating in each sequence.
What is a placebo?
People experiencing improvement in their condition even though they were assigned a dummy sugar pill without an active ingredients.
Transgenic mouse model?
The use transgenic knock-out or knock-in mice to investigate the independent variable.
What are the alternatives to full knockouts/ins?
- Conditional knockout – inactivate gene in specific brain regions
- Temporal conditional knockout – inactivate gene only at certain times
Terminology X and On
- X = intervention, treatment
- On = observation at time n
Between-subject designs non or quasi?
- Differential research design (non)
- Posttest-only nonequivalent control group design (non)
- Pretest-posttest nonequivalent control group design (quasi)
What is differential research design?
- Compare two groups of participants as they are
- Participants formed by their pre-existing variable
- Groups are nonequivalent even at the very beginning of the study
- No manipulation or intervention in the groups
What is are time-related variables?
Studies where observations are made over time
* These are called pre-post designs
* Recall within-group designs
But no counterbalancing to control order effects
Pretest must always come before the posttest measurement
Within-subjects design time-related variables? (4)
- One group pretest-posttest design (non)
- Time series design (quasi)
- Interrupted time series design (quasi)
- Nonequivalent control time series design (quasi)
Threats to internal validity for pre-post design? (2)
- History, maturation, instrumentation, order
- Experimental – counterbalancing to control order effects
What is One group pre-post design (non)?
Within-group, comparing results before and after treatment.
What is time series design (quasi)
Take multiple measurements before the treatment and multiple measurements after.
Time series design (quasi) disadvantages? (2)
- Time consuming, expensive
- History can still be a confound
Types of Developmental Research?
- Study change as a function of age
- Longitudinal design (within-subject)
- Cross-sectional (between-subject)
What is Cross-sectional design? (pros and cons)
- Pros: All individuals studied at the same time
- Cons: Different individuals at each age group
- Cons: Cohort effects – differences between age groups (cohorts) due to unique characteristics of that age Age 10 Age 15 Age 20
What is Longitudinal design? (pros and cons)
Cons: The same individuals are studied over time
* Takes a very long time
* Expensive
* High attrition
Pros: Compare between different ages therefore done in the
same individuals
Quasi-experiments often have greater external validity (true or false)
true
Quasi-experiments have reduced internal validity (true or false)
true
Descriptive Research goal?
Describe variables as they
exist naturally (NO RELATIONSHIPS/NO MANIPULATION)
Observational Research design?
The researcher observes and records behavior for the purpose of describing behavior.
Observational Research design strengths?
- Researchers observe behavior in real time
- High external validity
- Comprehensive: All features of the behavior can be observed
Observational Research design weaknesses?
*Ethical concern: spying on people
*Does not examine causes or relationships
What is a behavioral observation?
Involves directly observing and recording behavior in its natural setting.
Quantifying behavioral observation?
- Frequency method
- Duration method
- Interval method
What is Naturalistic observation or nonparticipant observation?
When a researcher observes behavior in its natural setting.
Strengths and weaknesses of Naturalistic observation or nonparticipant observation?
Strengths:
* Excellent for behaviors that cannot be manipulated for ethical reasons
*High external validity
Weakness:
* Time consuming
What is Participant observation?
When the researcher interacts and integrates with the participants to observe them.
Strengths and weaknesses of Participant observation?
STRENGTHS:
* Allows researchers to observe behaviour that is not normally open to scientific observation
* Allows the researcher to gain a unique
perspective about the study
* High external validity
Weaknesses:
* Extremely time-consuming
* Potentially dangerous
* The researcher could alter the participant’s behaviour
* Observer may lose objectivity
What is Contrived observation or structured observation?
When the observer sets up a situation that is likely to produce a desired behaviour.
Strengths and weaknesses of Contrived observation or structured observation?
STRENGTHS:
* Takes less time, researchers do not have to wait for behaviour to occur.
WEAKNESSES:
* Less external validity; because the environment is less natural the behaviour may be as well.
What is Content Analysis?
Measures the behaviours in books, movies, or other media
What is Archival Research?
Measures behaviour from historical records
What is Survey research design?
Uses a survey to obtain a description of a group of individuals.
What other kind of question?
1) open ended
2) Restricted questions
3) Rating-scale questions
What is non-response bias?
When a certain population does not respond to the survey (reduces external validity)
Kind of surveys?
- Internet
- Telephone
- In-person
What is Interviewer bias?
During contact with a participant, an interview can influence the participant’s natural responses.
What is a case study design?
A study of a single individual for the purpose of obtaining a description of that individual.
What is the idiographic approach to research?
The intense study of one individual.
Correlational research strategy graph?
- Relationship between variables is visualized in a scatterplot
- X-axis: Predictor variable
- Y-axis: Criterion variable
What are the Three characteristics of a relationship in a correlational research strategy?
- Direction
- Form
- Strength
Direction in scatterplot?
- Positive – both variables change in the same direction
- Negative – variables change in opposite directions
Strength in scatterplot?
- Indicated by correlation coefficient (+1.00 thru -1.00)
- Strongest correlation = +1 or -1
- indicates a positive relationship
- indicates a negative relationship
When to use Spearman or Pearson?
Pearson when linear and Spearman when monotonic. When cannot be used is when non-linear.
What is the Coefficient of determination?
r^2 -> when r^2=1 -> shared variance 100%
Percentage of change in one variable accounted for by changing the other variable = shared variance
Advantages of correlational research? (1)
- High external validity
Disadvantages of correlational research? (1)
- Correlation ≠ Causation
- Directionality problem