Karteikarten
PS295 2nd Half
Karten in diesem Stapel (98)
- 1Contingency TableCompares the patterns of frequencies
Depicts the number of people at each level of the IV and DV - 2How to compare in a contingency tableCompare the proportion of people in column with the number of people in column two who responded to that level...greater difference, stronger relationship
- 3Steps of research in correlational and experimental1. formulate hypothesis
2.operationally define the variables
3. Collect and summarize data
4. Interpret the Results - 4Interpretation/Explanation in correlational studyCan't be confident that a relationship is casual
Can discuss patterns with rationales - 5Why is it difficult to demonstrate causation?There are so many extraneous variables that are not controlled and confounds
- 6Extraneous VariablesAny variable that may have influence on IV or DV
We are not interested in these variables - 7ConfoundsFactor that varies with the IV
Specific type of Extraneous variable that is allowed to change with the IV and impact the DV - 8Third VariableConfounding variable that has a causal impact on the IV and DV
- 9Possible Causation Patterns1. Causation
2. Reverse Causation
3. Third Varaible - 10Understanding Control in Experiments: Control Over IVManipulation, creates levels, cause precedes effect
- 11Understanding Control in Experiments: Control over extraneous variablesHolding variables constant - consistency
Random assignment - balancing or equalizing - 12Control by holding variables constantMake sure everyone is exposed to the same extraneous variables : cover story, rater, room, target being studied
- 13Random AssignmentEvery member of the sample has equal opportunity to being chosen for the experimental condition
GOAL: to ensure they are roughly equal
randomizes subject influences - 14Types of random assignmentSimple Random Assignment &
Blocked Simple Random Assignment - 15Simple Random AssignmentGood with large number of participants, there could be different numbers of participants each group
- 16Blocked Simple Random SamplingSolves problems from simple random sampling. Define participants into blocks, assign each person in the block to the conditions = equal # of people in each condition
- 17Three Essential Properties of Experiments1. Manipulate the IV
2. assign participants that ensures initial equivalence
3. control extraneous variables - 18Three things needed to show causation1. Covariation - 2 varaibles must be related
2. Temporal Precedence - cause needs to come before effect
3. Control Over Extraneous Variables -addressed by holding constant and randomization - 19Possibilities of confounds in experimentsConfounds lower in experiments than in correlational studies
BUT if the researcher changes more than the IV there are confounds - 20Obvious ConfoundsOnes that you know are there
- 21Subtle ConfoundsAn issue in the construct validity
inadvertently change a different psychological process - 22Process Explanation In experimentsAttempt to identify the mechanism or mediating process
The process that intervenes between the IV and DV
How and why the IV exerts its effect - 23Internal ValidityDegree to which you can draw accurate conclusions - the difference in the DV are caused by the IV
Mostly in causal relationships - 24When is there higher internal validityThe IV was manipulated
The extraneous varaibles are controlled
No confounds - 25External ValidityThe degree to which findings can be generalized to other samples, settings and procedures
- 26Limits and Criticisms of External ValidityMay be tradeoffs between external and internal validity "the experimenters dilemma" - lab experiments are not representative and are not realistic
- 27Views of experimental researchersInternal validity is main goal over generalization
Goal to test hypothesis about causes - 28Unrepresented SamplesLab findings are not usually replicated in other settings
- 29Mundane RealismHow similar the study is to what you would encounter in real life
- 30Experimental RealismCaptures what you're trying to get at
- 31Quasi- ExperimentNot a true experiment - lack manipulation of IV or random assignment
Not as high in internal validity
Has design feature to eliminate threats - 32Threats to internal validity (Campbell and Stanley) - similar to confounds1. History
2. Maturation
3. Regression to the Mean
4. Pretest Sensitization (testing)
5. Instrumentation
6. Differential attribution
7. Selection - 33HistoryChanges on the DV produces by things other than the IV, something that happens between the pre and post tests, such as an environmental change outside of the study, can effect most or many of the participants
- 34MaturationNatural changes that occur to the participant as a function of time - biological or psychological, developmental changes, learning, boredom, hunger
- 35Regression towards the MeanChanges on the DV due to extreme scores. When participants are assigned, happens when people with extreme scores get assigned to a specific condition, their scores usually are normaler when retested
Worse with low reliability - 36What is regression towards the mean as a concept?Obtained scores = true score +/- error component
People with the same true ability can get different scores
People with different true ability can get the same scores - 37Misrepresentation of regressionSports illustrated jinx - when someone does really well they probably wont be that good again
- 38Summary of regression towards the meanregression will be different across groups
high and low groups will show regression in the opposite direction
average group will not show regression
shouldn't make groups based on scores - 39Pretest Sensitization (Testing)Completing a pretest can influence response on the posttest
Can sensitize participants to the hypothesis, they might think about the DV more, practice causes improvement, they might feel pressure to stay consistent - 40InstrumentationChanges on how the DV is assessed
Raters use different codings, change wording of questions, change in the measuring device
CAN record both sessions and mark them both in regards to how you would mark the first, then the second - 41Differential AttributionLoss of participants in the study
AKA mortality
those who drop out are different than those who stay, can make the groups different sizes - 42One-group pretest - posttest design1. Pretest DV
2. Introduce change on the IV
3. Posttest on DV - 43True Experiments (Randomized Group Designs)Use a control group that receives a different level of the IV
-Manipulation
-Random assignment to the groups - 44Solomon Four Group DesignCombination or pretest-posttest design
Allows to evaluate the possibility of a testing effect
E1 Pretest Treatment Posttest
E2 -------- Treatment Posttest
C1 Pretest ------------ Posttest
C2 ---------- ----------- Posttest - 45Additional Threats: Miscellaneous ConfoundsLots of confounds that can be introduced as part of the operationalization of the IV
- 46Experimenter ExpectanciesRosenthal Effect
Expectations effect their interactions with participants
Should keep the experimenter blind to which group is the condition - 47Participant ExpectanciesNotion of reactivity
Act different if they know what the test is about
Often want to behave how they think they should, how the experimenter wants them to - 48Demand CharacteristicsAny aspect of the research setting that subtly suggest how the participant should behave
SOLUTIONS: fake cover story, unobtrusive measures, make participants blind to their condition - 49Demand Characteristics - Cooperative ResponseThey try to act how they think they should
- 50Demand Characteristics - Uncooperative resposeThey try to undermine the research
- 51Placebo effectsSimilar to demand characteristics
Good to use a placebo in a control group
With treatment the participants think it will affect them - 52Double Blind ProcedureNeither the experimenter or the participant knows who is the control and who is the experiment
- 53Distribution of Scores in experimentsShape of the distribution tells us how much the IV affected the DV
- 54When there is a strong effect of the IV on distributionLittle overlap between the two distributions, appears bimodal
- 55When there is no effect of the IV on the distributionLots of overlap between the two distributions, appear to be one distribution
- 56Importance of the Concept of Variable in experimentsGoal of research is to understand variability... by manipulating the IV we can see the variability (changes) on the DV
We assess how much the total variability of the DV is due to the IV - 57Conceptual components of the total variabilityTotal variability on the DV = systematic variability + error variability
means = variability bc of the manipulation (and confounds) + variability due to extraneous varaibles with each condition - 58Statistical Components of total variability on the DVOverall Variability = between-groups variability + within-groups variability
Total variance of scores = difference between condition means + variance within each condition - 59Inferential Statisticsused to make inference about population based on sample data
decide whether the observed difference in means is greater than what we would expect error variance alone would be - 60Systematic Variability/Total Variabilitymeans
systematic variability (between groups)/ error variability (within groups) - 61Two situations that would create overlapping groupsthe IV did not affect people
the IV has affected people but extraneous variables within each condition work against the affect
CAN'T be confident in the results - 62Sources of VariabilitySystematic Variability and Error Variability
- 63Systematic Variabilitybetween groups
the IV manipulation
Confounds - 64Error Variabilitywithin groups
Extraneous varaibles such as:
-individual differnces, transient states, environmental factors, differential treatment, unreliable measures - 65Implications of the Types of Variance for Designing experimentsMake sure systematic variability is only coming from the IV (no confounds)
Make sure systematic variability is high (strong and weak manipulations)
Keep error variance to a minimum (eliminate extraneous varaibles) - 66Inferential Statistics and Hypothesis TestingIn experiments inferntial statistics helps us determine if the difference between means is greater than would be expected on the basis of error variance alone
- 67Hypothesis Testing: The t-testUsed for analysis of two groups experiments
test wether the difference in the two means is significant - 68Two steps of the t-test1. calculate an obtained t value
2. compare it to a critical t value (or calculate the probability of the obtained values) - 69The obtained t-value will be larger whenbetween groups variance is high
within group variance is low
sample size is larger - 70What percent probability does the t value usually have5% (found in a table)
- 71When are t value findings significantwhen the obtained t-value is larger than its critical t value (P < .05)
- 72More about t-valuesthey can be positive or negative
hypothesis can be directional or non directional
there is still a role of confounds - 73Hypothesis Testing: The Null HypothesisWe can decide if we can reject the null hypothesis
- 74Degree of freedomHow many people your studying minus the number of groups
- 75Alpha Level0.05, the amount it can occur (that you're willing to accept)
- 76Null HypothesisThere is no real difference between the groups
Any observed difference is due to the error variability alone - 77Statistical Hypothesis TestingAssign a probability level (p-level) to the statement that results are due to chance
Decide based on this p-level whether you reject the null hypothesis - 78Type 1 ErrorRejecting the Null hypothesis when it is actually true
-set a low p-level for rejecting the null (usually less then .05 - this is alpha level) - 79Type 2 ErrorFailing to reject the null hypothesis when it is false
Probability of making type 2 error (beta) - 80Power (statistical power)the ability to detect real differences
increases with larger sample sizes
power = 1 beta - 81Type 1 vs 2 which is worse?It is more important to avoid type 1 error
thus alpha is set very low - 82When to make a statistical decisionIf the observed t-value exceeds the critical t-value, then reject the null hypothesis (confirm ur hypothesis)
- 83Confidence IntervalsSample statistic is an estimaite of the true value in population
Usually set CI 95%
(95% probability the population mean will fall within the interval) - 84Effect SizeWant to know the strength of the discovered effects, how much of an effect did the IV have on the DV
- 85Effect Size Indicators1. Eta-squared and omega squared - amount of the total variance in the DV accounted by the IV
2. Cohen's D - size of the difference between two means relative to SD
3. Odds ratio - odds of an event occurring in one group compared to another - 86Types of experimental Designs-two dimensions that make up differences
If the same ppl or different ppl are in the two conditions (between/within)
If the design allows the researcher to asses how much subject variables are affecting the score (correlated/uncorrelated) - 873 Types of Experimental Designs1. Randomized Groups Design - between/uncorrleated
2. Matched-subject Design - between/correlated
3. Repeated Measures Design - within/correlated - 88Correlated Groups DesignResearcher tries to assess the influence of subject variables
Called this because when the SV is having on influence on the scores, the two conditions are correlated
SV correlates scores in different conditions - 89Subject VariablesIndividual Differences, qualities that the participant has before the study and brings in with them, relatively stable but may influence the DV
- 90Sources of within group varabilityindividual difference, transient states in participants, environmental factors, differential treatment, unreliable measures
- 91How to test for subject variable○ Step 1: identify a subject variable that we think would influence the scores within each condition (based on past research)
○ Step 2: prior to the study, for each member of the sample, obtain a measure of the subject variable that may influence scores on the DV
○ Step 3: list the scores on the subject variable from highest to lowest
○ Step 4: form pairs of participants that are as similar as possible
Step 5: Randomly assign the members of each pair (or block) to the conditions - 92Advantages of MatchingEnsures groups are equivalent on that variable, most useful when:
small sample
high variability can be measured
relevant variable can be measured ahead of time - 93Disadvantages to MatchingKnowing which variable to use
cost and effort - 94The Problem of order effectsmanipulation of the IV is confounded with passage of time,
fatigue, bordem, anxiety goes down,
testing effects - 95CounterbalancingDifferent participants face different orders of the conditions
effects associated with order will be balanced - 96Types of counterbalancingComplete Counterbalancing - every possible order
Incomplete Counterbalancing - latin square, only a subset of possible orders - 97Common numbering of conditions2 conditions - 2
3 conditions - 6
4 conditions - 24
5 conditions - 120
6 conditions - 720 - 98Building at Latin Square