Exam 2 study guide
Wednesday, November 16, 2016 1:04 PM
• External validity: the extent to which the results can hold true outside of the specific study ○ Can the results be generalized?
○ Would you get the same results in different conditions?
○ Types of generalization
▪ Generalization from a sample to the population- the sample needs to be representative of the population ▪ Generalization from one research study to another
▪ Generalization from a research study to the real world
○ Threats to external validity:
○ Generalizing across participants/subjects
▪ Selection bias- the sampling procedure favors the selection of some people over others
Don't forget about the age old question of A company issues 400 shares of %5 par value preferred stock for $12 per share. what amount will be credited to the preferred stock account when recording this transaction?
▪ Volunteer bias- volunteers arent perfectly representative of the population If you want to learn more check out Who benefits from alienated labor?
Cross-species generalizations- when a study is conducted with nonhumans, are the results really applicable to humans?
○ Generalizing across features of a study:
▪ Novelty effect- individuals may respond differently than they would in the real world
Multiple treatment interference- participating in one condition may have an effect that carries into the next treatment. Ex- fatigue, practice Don't forget about the age old question of What are the types of natural selection?
▪ Experimenter characteristics- such as personality, age, gender, etc.
○ Generalizing across features of the measures:
▪ Sensitization- the pretest may sensitize the participants so that they become more aware of their attitudes/behavior ▪ Generality across response measures- variables can be defined and measured in many different ways. Ex- fear ▪ Time of measurement- the effect of the treatment may change with time We also discuss several other topics like What are the steps in conducting scientific research?
• Internal validity: was there a cause and effect relationship?
○ Were changes in one variable followed by changes in the other variables?
○ No other variables can provide an alternate explanation
○ Allows only one explanation for the results
○ Threats to internal validity:
○ Extraneous variables- additional variables in research study that are not directly investigated
Confounding variables- an extraneous variable that changes systematically along with the two variables being studied. Usually can provide an alternate explanation for the relationship between the two variables
○ The only difference between treatment conditions should be the single variable that was used to define the conditions ○ Environmental variables
○ Individual differences- personal characteristics that differ from one person to another
○ Assignment bias
○ We also discuss several other topics like Give a function of carbohydrates.
Time related variables: same group of participants, their scores are compared over time. During the time between the conditions, the participants can be influenced by other factors. Ex- fatigue
• Balancing internal and external validity:
High internal validity means minimizing confounding variables, must tightly control the experiment. This creates an artificial environment that affect the results
○ High external validity means that the research environment closely resembles the real world • Artifacts- an external factor that may influence or distort measurements, threatens internal and external validity
Experimenter bias- experimenter's expectations or beliefs of the outcome influence the findings of the study. Single blind and double blind studies minimize experimenter bias.
• Exaggerated variables- researchers tend to exaggerate the differences between conditions in a study • Random error
○ Can be in the instrument or the person being measured
○ Cancels out with repeated measures
○ Can be due to intrinsic noise- ex: drinking a lot of water before being weighed
○ Can be due to observer error. Ex- reading from a scale
• Systematic error Don't forget about the age old question of What is documentary photography?
○ Consistent error
• Measuring reliability:
○ Inter-rater reliability- same observations, different raters. The correlation between the ratings of different judges ○ Test-retest reliability- same measurement taken at different times. The correlation between a test at different times/trials
Split-half reliability- same test, different items assess the same things. Within the same measure. Randomly splits the test into two halves, correlate the scores against one another. The results will be positively correlated if the test is reliable
Exam 2 study guide Page 1
two halves, correlate the scores against one another. The results will be positively correlated if the test is reliable • Construct validity- does the experiment measure what the research claims it does?
○ Are the operational definitions reasonable measures of the construct?
○ Convergent validity- show that 2 measures of the same construct are correlated
Divergent validity- show lack of correlation between the measure of interest and a different measure. Want to rule out the possibility that the operational definition is measuring a different construct
• Extraneous variable : any variable besides IV or DV
• Confound: an EV that systematically varies with the manipulated variable and explains the data ○ Ex- coke vs pepsi label with S or L
• Want to cause a change in the dependent variable, want to get rid of confounds • Directionality/ temporal precedence: which comes before the other
• Elements of experimental design:
○ Manipulation- create treatment condition. Levels of IV
○ Measurement- a variable that should change- scores of a DV
○ Comparison- scores on DV across levels of IV
○ Control- eliminate confounds
Holding constant- hold the value of a potential confound constant across all levels. All items have the same value or a restricted range. Ex- all participants the same age/ range
○ Ex- does having big feet cause kids to read better?
○ IV: shoe size. Levels- size 4 vs size 6
○ DV- reading test
○ Confound- age
• Matching- for every level of IV there exists one item or person with the same value of the potential confound ○ Ex- size 4: 5 kids age 5,6,7,8. size 6: 5 kids age 5,6,7,8
○ Tightly controlled way to cancel out differences in a potential confounding variable
Random assignment: how we place the sample into the conditions of the experiment (Levels of the IV). Everyone has an equal chance of being put into one of the conditions
○ Not possible for shoe size and age, because the younger kids tend to have smaller feet and tend to be in that group ○ Any differences should be equally spread across conditions by chance and should cancel out
• Matching vs random assignment:
○ Matching best when you have a small number of participants, and concern is about 1 potential confounding variable. ○ Random assignment best when there are lots of participants and not concerned with 1 specific extraneous variable • Control groups: help determine the effect
• Placebo effect- due to expectation alone
Between vs within subjects designs:
• Between subjects designs
○ Aka independent measures
○ One level of IV per group of people
○ DV is the same for each group
○ Ex- 1 group gets drug, the other gets placebo
▪ Clean measures
▪ No practice or fatigue effects
▪ Can ask WHO questions about individual differences
▪ Needs lots of resources and more participants
▪ Can have differences in participants not related to IV-- individual differences and potential confounds • Sometimes hypotheses are about differences between people
○ Quasi-experimental study/individual difference study
○ Exploits differences that naturally exist
○ Male vs female, young vs old, etc.
• Controlled difference study
○ Attempts to eliminate or minimize differences between groups
○ Uses experimental control--matching, holding constant, random assignment
• After only design- only measuring the DV after the treatment
○ Uses random assignment
• Before-after design/ pretest-posttest design: test before and after the IV manipulation. Measure difference and compare across Exam 2 study guide Page 2
Before-after design/ pretest-posttest design: test before and after the IV manipulation. Measure difference and compare across groups
○ Uses random assignment
• Experimenter bias
If experimenter is aware of groups, they may inadvertently influence the results through tone of voice, gestures, reinforcing desired behaviors, misinterpreting behavior.
○ Use single or double blind study
• Between vs within vs mixed design
○ Between- each participant receives 1 level of IV
○ Within- each participant receives every level of the IV
○ Mixed- multiple IVs, at least 1 is between and 1 is within
• Within subjects/repeated measures design
○ Each participant is in every condition, receives every level of IV
○ Advantages- no worries about differences between groups, statistically more powerful
Disadvantages- prior behavior or decisions may affect later ones, changes over time, test effects like fatigue, carryover effects, attrition (dropping out)
• Counterbalanced design-
Random method for counterbalancing: subjects split randomly into groups. Used when there are many conditions. Each subject gets the conditions in a random order
Balanced- an equal number of subjects receive each order. Ex- run 1 mile in blue shoes and 1 mile in red shoes. Another group runs in red shoes first then blue shoes.
• Parallel lines- at least 1 main effect, no interaction
• V shape graph- at least 1 main effect, plus interaction
• X shape- only interaction, no main effects
• Used for more complex, real life situations, because behavior is influenced by a variety of factors acting together • Very powerful, can be used for any combination of factors
• Can see how each individual factor influences behavior and how the group of factors acting together influence behavior • Include more than 1 independent variable
• Factors- when 2+ independent variables are combined in a single study, the independent variables are called factors • Main effect- the mean differences between columns. Statistical test needed to see if the mean difference is significant • Interaction- occurs when two factors produce mean differences that are not explained by the main effects of the two factors ○ One factor has a direct influence on the effect of a second factor
○ Exists when the effects of one factor depend on the different levels of a second factor. They are interdependent ○ Ex- drug interactions when one drug influences the effects of a second drug
• Principles for ethics in research
○ Nonmaleficence and beneficence
○ Fidelity and scientific integrity
Autonomy- the participant must have the unbiased/uncoerced right to know what they are participating in and to decide whether or not to participate.
○ Informed consent ensures autonomy
○ Withholding of incentives is coercion
○ Violation of autonomy- Jewish chronic disease hospital,1963
▪ Hypothesized that debilitated pts will reject foreign cells
▪ 22 chronically ill patients who did not have cancer were injected with live human cancer cells
Physicians did not inform pts what they were doing because they did not want to scare them, and they thought the cells would be rejected
• Trust- ensure a relationship of trust between the participant and researchers ○ Confidentiality- your identity is linked to your data, but we keep that link secret ○ Anonymity- your identity is not liked to your data
○ Deception- use only when absolutely necessary
○ Debriefing- educating participants about the design and purpose of the experiment • Justice- burdens and rewards are distributed equally.
Exam 2 study guide Page 3
• Justice- burdens and rewards are distributed equally.
IRBs and grant agencies often require researchers to record the race and gender of participants to ensure it matches the demographics of the population.
○ Must justify using any other distribution
Nonmaleficence and beneficence- to avoid bringing harm to research participants and take steps to maximize the benefits of research. Minimize the risks.
○ Benefits must outweigh the costs
• Fidelity and scientific integrity- researchers must be honest ○ Misconduct/fraud- making up the data, leaving out relevant data ○ Goal of peer review is to ensure fidelity and scientific integrity
Diederick Stapel: studied prejudice, published 150+ articles, he admitted that he never ran experiments for some papers, fabricated the data, lied to colleagues and students
• Oversight in the US
○ People- IRB- Institutional review board. Confirms that the research meets an acceptable ethical standard ○ Animals- IACUC
• Animals in research:
○ APA guidelines
▪ Qualified and trained individuals
▪ Research must be justified
▪ Must minimize harm and discomfort
○ 3 R's of animal research
▪ Reduce the number of animals used
▪ Refine to cause least stress
▪ Replace animals with other models
• Major ethics violations occurred in the Tuskegee syphilis experiment
Exam 2 study guide Page 4