Home
Subjects
Explanations
Create
Search
Log in
Sign up
Upgrade to remove ads
Only $2.99/month
Social Science
Psychology
Experimental Psychology
Psych 12 Midterm 3
STUDY
Flashcards
Learn
Write
Spell
Test
PLAY
Match
Gravity
Terms in this set (73)
Bivariate correlation
A statistical relationship between exactly two variables
- measure two variables (either through self-report, observation, or physiological measure) in the same group of people
-most appropriate when both measures are continuous (ordinal, interval, or ratio)
How are correlations visually presented?
scatterplots
what test do you use with a bivariate associations with one continuous and one categorical?
independent samples t test
Independent samples t-test
another way to analyze the relationship between exactly two variables
- appropraite when one variable is categorical (nominal)
- visually represented with bar graphs (usually) and scatterplots (sometimes)
a correlational study
one in which all variables are measured and none are manipulated
(if manipulated then its a CASUAL CLAIM)
What are the five questions for statistical validity of correlations?
1) Effect size (r value)
2) Statistical significance (p value)
3) Outliers
4) Restriction of Range
5) Curvilinear associations
Question 1:
What is the effect size?
The strength of a relationship. In a scatterplot, the tighter the data points are to a line, the stronger the effect
- r statistic: ranges from -1 to 1
- direction: positive, negative, or zero
- closer to -1 or 1= stronger correlation
- closer to 0 (-.03, .06)= weak or no correlation
How strong is our effect based on the r?
- .1 (or -.1) = small/weak
- .3 (or -.3)= medium/ moderate
- .5 (or -.5) = large/ strong
Why do we care about effect sizes?
they give us knowledge to make more accurate predictions
Larger effects (are usually more important because
Large effect sizes+ more accurate predictions
-strong correlation means strong prediction
Small effect sizes can?
still be important (ex: aspirin study)
Statistical significance (p value)
the likelihood that the effect would come out that strong by chance, assuming no effect in the real world
- we generally tolerate up to 5% possibility that the effect occurred due to chance
p-value
probability that out sample came from a population with zero association
- p-value less than .05 is considered 'significant', greater than .05is 'nonsignificant' or 'not significant'
P values are influenced by what?
effect size and sample size
outlier
an extreme score on either or both variables
why outliers are a problem?
- they can exert a strong, disproportionate influence on the size of the correlation
- can make correlation much stronger/weaker than it would have been
- can make the correlation significant/ non-significant because significance is related to effect size
Restriction of range
when you only have part of the full scale for one or more of your variable represented in sample
-artificially reduces the strength of a correlation
- can also be misleading (for the question; is the relationship curvilinear)
Curvilinear relationship
a relationship that is not well-represented by a straight line (not linear)
-normal correlation (r) only tests for linear relation
- variables might be associated in nonlinear ways
- correlation will be weak (attempts to fit a straight line to the data), but could still be interesting
What are the requirements for causation?
1) covariance: as one changes, the other also changes
2) temporal precedence: cause must precede effect (directionality)
3) internal validity: nothing else can explain the relationship between two variables (third variable)
Multivariate designs
correlational studies that involve more than two variables
Longitudinal designs
measuring the same variable(s) repeatedly at several points in time
1. research example: measured optimism and anxiety at 9 points in time
-once before participants took the bar exam
- eight times during the waiting period (every 2 weeks)
Multi-regression analyses
predicting an outcome from more than one predictor variable to narrow down the relationship with he predictor on interest
cross-sectional correlations
are the two variables correlated within the same point in time?
-just like if you only measured the variable at one point in time
autocorrelations
is each variable related to itself across time
-similar to establishing test-retest reliability
cross-lag correlations
is the earlier measure of the variable associated with the later measure of the other variable
- addresses temporal precedence (reduces directionality problem)
- if one cross-lag correlation is stronger than the other, suggest that one variable precedes the other
-also possible that both cross-lag correlations are strong (or weak), and then temporal precedence is not established
multiple-regression analyses
predicting an outcome from more than one predictor variable to narrow the relationship with the predictor of interest
- goal: attempt to rule out third variable (increasing internal validity)
controlling
means eliminating the effects of a 3rd variable to look at the unique contribution of the predictor
-holding a potential third variable at a constant level while investigating the association between two other variables
criterion variable
the variable you are trying to predict; similar to dependent variable
predictor variable
the variable that might be causing change in criterion variable; similar to independent variables
what can solve internal validity
measuring every potential confound and statistically control for their effects on the outcome
- however it is not possible to measure every confound but regression helps eliminate some
"controlling for"
holding a potential third variable at a constant level while investigating the association between tow other variables
Beta
statistical representation of the relationship between each predictor variable and the criterion variable
-similar to an (r) in a bivariate correlation : ranges from -1.0 to +1.o, has an associated p-value
-negative beta= negative association
-positive beta= positive association
- beta= 0 means no association
-larger positive (.2 compared to .1) or larger negative (-.2 compared to -.1) βs indicate a relatively stronger association for one predictor compared to the other
beta has a p value that indicates statistical significance
- how likely is the association due to chance?
- p < .05 suggest there is truly an association even after controlling for other variables
Mediation
tells us why two variables are associated
mediator
a variable that explains the relationship between two other variables
-answers the WHY question
-testing mediation involves correlations and multiple-regression
Four steps to establish mediation (simple method)
1. test relationship A to C (called path c)
2. test relationship A to B (called path a)
3. Test relationship B to C (called path b)
4. Multiple-regression: does c hold up after controlling for the mediator? (c')
** if c' stays the same as c (from Step 1), then B is not a
statistically significant mediator. If c' is smaller than c, then B is a mediator
moderator
asks when, or for whom, are two variable associated
mediators vs confounds
mediator explains why two thins are casually related, whereas a third variable that is a confound explains why two thins seem causally related but really aren't
manipulated variable
a variable that is controlled by the researchers, who assign participants to experience particular levels of the variable (independent v)
measured variables
records of thoughts, feelings, or behaviors, not directly influenced by the researchers (dependent v)
control variables
any variable the researcher intentionally holds constant across conditions
independent variable
the manipulated variable in an experiment
conditions
the levels or version of the independent variable
dependent variable
the measured variables in an experiment
what are the two ways to check construct validity of manipulated variables
1) pilot study
2) manipulation check
Pilot study
conducted before the "actual" study to check the construct validity of a manipulation
manipulation check
an extra measure designed to see how well a manipulation worked
types of conditions
1) control group
2) treatment group
3) placebo group
control group
a condition that is supposed to represent "no treatment" or a neutral state (business as usual)
treatment group
the condition of interest, which are compared to the control group
placebo group
a control group who believes they're a treatment group, with the goal of ruling out expectancy effects
confound
anything that differs between your groups OTHER THAN the levels f the independent variable
design confounds
something that inherently varies along with the independent variable
-creates systematic variability (big problem)
-unsystematic variability: created when something differs among participants but does NOT systematically co-occur with the independent variable
random assignment
each participant has an equal chance of being in each condition
- maximizes the likelihood that unintended variability is unsystematic instead of systematic
-avoids selection effects
a true experiment has two essential characteristics, which are?
1)Manipulation of one or more independent variables
2) random assignment to conditions
selection effect
when the kind of person in one condition are systematically different from the ones in other conditions
important concept: "chance is lumpy."
- randomness doesn't always end up seeming random
- can create a "failure of random assignment"
matched groups (or matching)
ensuring that your groups are equivalent in important ways
- pair (match) people on the characteristics if interest ,then split the pair across conditions through random assignment
between-subjects design
each participants is only in one experimental condition (independent-groups designs)
two types of between-subjects designs
1)posttest- only design
2)pretest-posttest design
posttest-only design
participants undergo the manipulation (just one condition) and then complete the measures (once)
pretest-posttest design
participants first complete the measures, then the manipulation, then the measures again
pretest-posttest design advantages
- test and control for selection effects
- test and control fo failures of random assignment
pretest-posttest design disadvantages
-might create demand characteristic
-people might think they should be consistent in their responses
within-subjects design
each participant is in al experimental conditions
two type:
1)concurrent measures
2)repeated-measures design
concurrent measures
participants experience all levels of the independent variable at once
repeated measures design
participants experience levels of the independent variable one after the other, with the measures following each level of the independent variable
why would you want a within-subjects design?
1) guarantees equivalence of groups
(no selection effects)
2) functionally doubles your sample size (for two conditions)
- statistical power: ability of a study to get a statistically significant effect, assuming the effect is real
-remember, p values= function of effect size + sample size
why would you want between-subjects design?
- order effects
order effects
a confound that occurs when experiencing one condition changes how participants react to subsequents conditions in a within-subjects design
types of order effects
-practice effects: participants get better at measures
- fatigue effects: participants get worse at the measures
- carryover effects: effects of one condition contaminate subsequent repossess
- sensitization effects: participants become suspicious or clued in from earlier conditions
how do you deal with some order effects?
counterbalancing: randomly assigning participants to experience the conditions on different orders
types of counterbalancing
1) full counterbalancing: all possible orders are represented
2) partial counterbalancing: only some orders are represented
counterbalancing does't FIX order effects, it just allows you to check for them
Recommended textbook explanations
Psychology: Principles in Practice
Spencer A. Rathus
1,024 explanations
Myers' Psychology for AP
2nd Edition
David G Myers
900 explanations
A Concise Introduction To Logic (Mindtap Course List)
13th Edition
Lori Watson, Patrick J. Hurley
1,242 explanations
Bundle: A Concise Introduction To Logic Aplia 1 Term Printed Access Card
12th Edition
Patrick J. Hurley
1,108 explanations
Sets with similar terms
302 exam 3
50 terms
Morling: Research Methods in Psychology - Carroll…
96 terms
Intro to Research: Chapters 8-10
62 terms
Research Test 3 Review
78 terms
Other sets by this creator
Soc 001 9
33 terms
Soc 001 8a & 8b
48 terms
Soc 001 Review Terms for Midterm 2
6 terms
Soc 001 Midterm 2 Readings
6 terms
Other Quizlet sets
Transcendentalism
15 terms
TExES 113 Competency 010 (History) - Part 3: World
93 terms
Models of the solar system
18 terms
PT Anatomy & Physiology 102.4-102.8
50 terms