Search
Browse
Create
Log in
Sign up
Log in
Sign up
Upgrade to remove ads
Only $2.99/month
Research Methods of Psych 2
STUDY
Flashcards
Learn
Write
Spell
Test
PLAY
Match
Gravity
Key Concepts:
Terms in this set (64)
Developmental Research Methods
Used when interested in changes that occur with age/maturation
--Cross-sectional method
--Longitudinal method
--Sequential method
Cross-sectional method
Researcher studies people of different ages at single point in time
Advantages:
Relatively cheap (compared to other developmental designs); provides quick results
Disadvantages:
Possible confounding of age and cohort effects
Longitudinal method
Research studies people of one cohort at several points in time
"Cohort" doesn't have to be as broad as Baby Boomers or Gen X, Y, or Z
Examples
Snowdon's "Nun Study"
Dunedin Study
Advantages:
Authentic—allows us to study real changes occurring in real people as they grow older
Disadvantages:
Expensive; doesn't yield quick results; introduces concerns about study attrition and mortality; instrumentation concerns; potential confounding of age and time of measurement effects
Sequential method
Sometimes referred to as cross-sequential
Researcher studies people from multiple cohorts at multiple points in time
Example
Framingham Heart Study
Advantages:
Yields data quickly and over long haul; allows us to disentangle age and cohort effects, age and time of measurement effects
Disadvantages:
Most expensive of developmental designs; introduces attrition/mortality concerns, instrumentation concerns
Age effects
Consequences of being a certain age
Cohort effects
Consequences of being member of particular cohort
Cohort: Group of people born around the same time; exposed to the same historical events, cultural forces, etc.
**confound: age and cohort effects
Time of Measurement Effect
•Results from events or trends occurring around the time of data collection
•(Affects more than just one particular cohort)
Single-Case Designs
Advantages of Single-Case Designs
-Possess "individual-subject validity"
-Possible when
-Potential participants rare/hard-to-find
-Procedures super time-consuming or expensive
Elements of Single-Case Designs
1) Operationally define target behavior
2) Establish baseline for participant
3) Institute treatment
4) Continue to monitor behavior
Like an interrupted time-series design with an N of 1....
(Two) Types of Single-Case Designs
•Reversal designs
•Multiple baseline designs
Reversal designs
•Simplest design: ABA
A = baseline period, B = treatment period
•May be extended
-ABAB
-ABABAB
-etc.
•The introduction and withdrawal of treatment (especially more than once) allows one to rule out influence of history or maturation
•But reversal designs not always possible (practically, ethically)!
Multiple Baseline Designs
•Across
-Subjects
•Observe one behavior in several individuals over extended time
•Introduce treatment at different time for each
•If treatment followed by change in behavior in each case, treatment likely effective (rules out historical effects, e.g.)
-Behaviors
•Observe several behaviors in one individual over extended time
•Apply "same manipulation...to each of the behaviors" at different times (C&B, 2015, p. 224)
•If rate of each behavior changes after introduction of treatment, treatment likely effective
-Settings/situations
•Observe behavior of one individual across multiple settings
•Over time, apply treatment in one setting after another
•If rate of each behavior changes after introduction of each treatment, treatment likely effective
Criticisms of Single-Case Designs(from Goodwin, 2005)
1)Generalizability of results questionable
2)Evaluation of results may rely on "the mere visual inspection of the data" (Goodwin, 2005, p. 380) rather than statistical analyses
3)Not easy to test for interaction effects
4)Primary dependent variable = response frequency
Scales of Measurement
1)Nominal
The categories of a nominal scale simply name, or classify, observations
Honda, Ford, Volvo
Catholic, Jewish, Muslim, Southern Baptist
(Names only; provides no quantitative information)
2) Ordinal
The categories of an ordinal scale are ranked in terms of magnitude, so one can tell the relative standing of observations
Win, place, show in a horse race
(Names and ranks)
3) Interval
The categories of an interval scale form a series of intervals that are all exactly the same size
Fahrenheit temperature scale
IQ score
(Names, ranks, and has equal intervals)
4) Ratio
•The ratio scale has all the properties of the interval scale, plus a true zero point
Weight
Years of age
(Names, ranks, has equal intervals, and has true zero point)
Measurement (Redux)
Categorizing something that varies, or assigning it a numerical value
Three General Ways Psych Results May Be Described
1) Comparison of percentages/proportions of groups
e.g., Percent of 1st graders who have started to read independently, School A vs. School B
e.g., Percent of stay-at-home caregivers of each sex
2) Correlation between sets of scores
May be observed by looking at a scatterplot/by calculating a correlation coefficient
e.g., relationship between body satisfaction and self-esteem
3) Comparison of group means
•e.g., SAT scores of people who took no test preparation class vs. Kaplan vs. Princeton Review
Descriptive Statistics
•Measures of central tendency
•Measures of dispersion
Measures of Central Tendency
1)Mean
•Arithmetic average
How/why is the mean useful?
What is one big problem with the mean?
2) Median
•Middle score of (an ordered) distribution; score at the 50th percentile
•Less susceptible than the mean to the influence of outliers
•What is a median split?
--The process of using the median to transform continuous data into dichotomous data
3) Mode
•Most frequent response in a distribution
•Let us see how the average person answers the question, performs, etc.
•Let us determine whether ceiling or floor effects
•Only meaningful for certain data types
Median Split
The process of using the median to transform continuous data into dichotomous data
Here's a hypothetical distribution of BDI (Beck Depression Inventory) scores:
4, 8, 11, 11, 12, 15, 18, 26
What's the median in this case?
What scores would be assigned to the low depression group?The high depression group
Measures of Dispersion
1) Range
Difference between the lowest and highest values observed in the sample
Range of following distribution of test scores?
60, 80, 85, 95, 100
Advantages of range?
Disadvantages?
2) Interquartile range
Difference between the 25th and 75th percentiles
Advantages?
Disadvantages?
3) Variance
The average squared deviation of each score from the mean
4) Standard deviation
Square root of the variance
•Another key aspect of the data; measures of central tendency and dispersion both are needed to evaluate "statistical significance"
Double-Barreled
•Ask about only one thing (attribute, behavior, belief, attitude) per question!
•Better to have more simple questions than fewer complex ones
Leading/Loaded
•Leading—suggests the answer that the questioner wants confirmed
•Loaded—characterized by one or more implicit assumptions (typically controversial or unjustified)
Keep the language of questions (and response options) neutral!
Example of this is "Was your waiter friendly and fast?"
Inferential Statistics
Inferring characteristics of the population based on data from a sample
Type I, or alpha, error
rejecting the null hypothesis when null is, in fact, true
types of hypotheses
•What is the null hypothesis?
-H0: μ1 = μ2
•What is the alternative hypothesis?
-H1: μ1 ≠ μ2
-H1: μ1 > μ2 first alternative
-H2: μ1 < μ2 second alternative
•Hypotheses are constructed to be mutually exclusive and exhaustive
Alpha level or level of significance
probability value that is used to define very unlikely outcomes in the case where the null hypothesis is true
•By convention, alpha usually set to .05, .01
•If our p-value is smaller than .05, we do what?
-Reject the null hypothesis
•If our p-value is > .05, we do what?
-Fail to reject the null hypothesis
•The null hypothesis is never "accepted"!!!
Concerns with Hypothesis Testing(from Cohen, 1994, and elsewhere)
1) The result is an all-or-nothing decision with an arbitrary cutoff
2) The concept of the null hypothesis is inherently artificial ("Nil hypothesis"?)
3) People often misinterpret decisions; with NHST, we are able to conclude
-probability of data given null, not probability of null given data
Effect Sizes
•A variety of effect size measures; for example
-r
-Cohen's d
-eta2
-odds ratio
etc.
Confidence Intervals
•The confidence interval (CI) indicates how accurately the sample value describes the population value
e.g., the 95% CI is the interval around the sample value where we are 95% certain the population value falls
Parametric vs. Nonparametric Tests
Parametric Tests:
--Test hypotheses about population parameters
--Require that a number of assumptions about population distributions be met
--Require interval or ratio data
Nonparametric Tests
--Make few assumptions about population distributions
--Rely on ordinal or nominal data
--Not as powerful as parametric tests
Common Parametric Tests
1) t-test
2) One-way analysis of variance (ANOVA)
3) Analysis of covariance (ANCOVA)
4)Factorial ANOVA
Common Nonparametric Test(s)
1) Chi-square test for goodness of fit
2) Chi-square test of independence
Tests of Association
1) Pearson product-moment correlation coefficient (r)
2) Spearman rank-order correlation coefficient (rs)
3) Partial correlation
4) Linear regression
5)Multiple regression
t-test
Used to test for a difference between two means
•May be used in between-subjects designs (independent groups t-test) or within-subjects designs (dependent-groups t-test or paired t-test)
•One independent variable; IV (or QV) must be categorical, DV must be continuous
The appropriate test statistic to use when comparing the means (not percentages) of two groups
dependent t: The appropriate test statistic to use when comparing two scores collected from the same group of people (e.g., "before" and "after")
ANOVA
•Used to test for a difference between two or more means
•May be used in between-subjects designs (single-factor ANOVA) or within-subjects designs (repeated-measures ANOVA)
•One independent variable; IV (or QV) must be categorical, DV must be continuous
•Significant omnibus F-statistic indicates significant difference between at least two means
(The appropriate test statistic to use when comparing the means of two or more groups, or when analyzing a factorial design)
•Post-hoc tests necessary to determine where the differences lie
ANCOVA
•Like an ANOVA, but allows you to examine the effects of one IV/QV on a DV, controlling for the effects of another variable
Factorial ANOVA
•Used when interested in the effects of at least two factors (IVs or QVs), or the interaction of these factors, on a DV
•Again, IVs (or QVs) must be categorical, DV must be continuous
•Separate F-statistics are computed to test for main effects and interactions
•If significant main effects, must do post-hoc comparisons
•If interaction significant, must analyze the simple main effects; if interaction significant, this result receives more attention than the significant main effects
Chi-squared goodness of fit
•Used when the data are categorical, and we want to test hypotheses about the shape or proportions of a population distribution
•Involves computation of the chi-square statistic
•Significant chi-square value tells us that observed frequencies in sample differ from expected frequencies
The appropriate test statistic to use when comparing percentages
Chi squared test of independence
•Used when the data are categorical, and we want to determine whether or not our variables are independent
•Involves construction of a matrix, computation of the chi-square statistic
•A significant chi-square value tells us that the two variables are related (i.e., not independent)
•Follow-up testing may be required, given the size of the matrix
The appropriate test statistic to use when comparing percentages
Pearson product-moment correlation coefficient (r)
•Designated r
(The appropriate test statistic to use when measuring the linear relationship between two variables)
•Used when interested in exploring the relationship between two variables
•Used when both variables measured on the interval or ratio scale
Spearman rank-order correlation coefficient (rs)
•Designated rs
•Used when interested in exploring the relationship between two variables
•Used when one or both variables measured on the ordinal scale OR when the distribution for one or both variables is not sufficiently normal
partial correlation
•Used when want to determine relationship between two continuous variables while controlling for the effects of one or more additional variables
linear regression
Closely related to correlation
•Allows us to predict scores on a DV given a certain level of IV
multiple regression
•Allows us to predict scores on a DV given certain levels of more than one IV
Type II, or beta, error
Failing to reject the null hypothesis when null is, in fact, false
Remedies for hypothesis testing issues
1)Explore data graphically (e.g., by using Tukey's Exploratory Data Analysis, or EDA)EDA: box and whisker and stem and leaf display
2) Report effect sizes, and confidence intervals, in addition to indicating whether effect is "significant"
3) replicate replicate replicate!!
alpha
A measure of the internal consistency of a questionnaire
p-value
The letter used to represent statistical significance
percent agreement
A measure of inter-rater reliability
mortality
The dropout rate in an experiment
maturation
Threat to internal validity in a within-subjects design because of naturally occurring processes within participants (e.g., immune system)
significance
Probability of obtaining a test statistic as extreme or more extreme than yours just by chance
demand characteristics
Features of a study that can reveal the hypothesis to participants
selection effects
Threat to internal validity caused by pre-existing differences among levels of the independent variable
within-subject
Design in which the same participants are exposed to more than one level of the independent variable
power
The probability of obtaining statistical significance, given a known sample size and effect size
discriminant
Type of validity showing that a measure is distinct from related measures
random assignment
Technique designed to reduce the risk of selection effects
double-blind
Technique designed to reduce the risk of experimenter expectancy effects
interaction
Occurs when the effect of one independent variable is different at different levels of another independent variable
history effects
Threat to internal validity in a within-subjects design because of an external event occurring between measures
curvilinear
Type of relation between variables that is not captured by correlation
random sampling
Technique designed to increase external validity
mixed design
Design that contains both within-subjects and between-subjects independent variables
statistical regression towards the mean
Threat to internal validity occurring in a within-subjects design when a sample is selected because of its extreme scores (either high or low).
YOU MIGHT ALSO LIKE...
17
Psychology | Sdorow, Rickabaugh, Betz
Research Quiz # 2 - Quantitative Research
31 terms
Comps RESEARCH Design and Evaluation
84 terms
Statistics - Dr. Lai
85 terms
OTHER SETS BY THIS CREATOR
Psychology 102 Study Guide 3
47 terms
Psych 102 Second Exam
94 terms
Research Methods of Psych
172 terms
Econ Equations to know
56 terms