Use LEFT and RIGHT arrow keys to navigate between flashcards;
Use UP and DOWN arrow keys to flip the card;
H to show hint;
A reads text to speech;
113 Cards in this Set
- Front
- Back
- 3rd side (hint)
6 Steps for conducting an experimental research study |
1. Develop a hypothesis
2. Choose a research design 3. Select a sample 4. Conduct the study 5. Analyze data 6. Report results |
|
|
Protocol Analysis
|
A type of content analysis when experimenters ask S to "think outloud."
|
|
|
Interval Recording
|
Method for behavior sampling during discrete intervals (i.e., is it occring now). Good for sampling complex beh. with no clear cut beginning or end such as laughing, talking, or playing.
|
|
|
Event Sampling
|
Recording each time the event occurs. Good for beh. that infrequently happen.
|
|
|
Sequential Analysis
|
Coding behavioral sequences rather than isolated beh. events when studying complex social behaviors.
|
|
|
Situational Analysis
|
Alt to beh sampling and used when the goal is to observe beh in multiple settings.
|
|
|
Nonexperimental Research
|
Conducted to collect data on variables.
|
|
|
Experimental Research
|
Conducted to test hypotheses about the relationship between variables. (True exp or Quasi exp)
|
|
|
Variables
|
Characteristics or behaviors that researchers can vary.
|
|
|
Random Assignment
|
"Randomization" helps ensure that any observed diff between groups is due to the IV. Random assignment of S to control or experimental group.
|
|
|
Quasi-Exp Research
|
Must use intact pre-existing groups or a single treatment group. No random ass b/c your just using one group.
|
|
|
Random Sampling
|
Every member of the pop has an equal chance of being included. Reduces biased sampling.
|
|
|
Stratified Random Sampling
|
Dividing the pop into the "strata" (e.g., SES, ed., gender, age, ract)and then using random sampling.
|
|
|
Cluster Sampling
|
Selecting pre-existing units/clusters/groups of ind. Used when it's not poss to id an entire pop.
|
|
|
Random Assignment
|
Allows an investigator to be more certain that the DV was caused by the IV.
|
|
|
Random Selection
|
Enables the investigator to generalize findings from the pop to the sample.
|
|
|
Extraneous (confounding) variable
|
Source of systematic error that effects the DV, but is irrelevant to the research.
|
|
|
Techniques to control confounding variables:
|
1. Random Assignment
2. Holding the Ext Var Constant 3. Matching S's on the Ext Var 4. Building the Ext Var into the study ("Blocking") 5. Stasticial Control (ANCOVA) |
|
|
Random Error
|
Experimental research attempts to minimize fluctuations in S's, conditions, and measuring instruments.
|
|
|
Internal Validity
|
Successfully determining if there is a casual relationship between IV and DV.
|
|
|
8 Threats to Internal Validity
|
1. Maturation - S's change
2. History - External events 3. Testing - Learning tests 4. Instrumentation - Var in testing specificity (Ex: raters accuracy) 5. Stastical Regression - When extreme groups are studied 6. Selection - Really assignent problem. Systematic diff b/t groups at beginning of study. 7. Attrition - drop outs 8. Interactions w/ Selection - Int w/ history. One selected group is diff on a var. Nonequivalent groups. |
|
|
External Validity
|
Being able to generalize findings to other settings.
|
|
|
Threats to External Validity
|
1. Int b/t Testing and Treatment - Pretest sensitization
2. Int b/t Selection and Treatment - Pre S variables such as motivation of volunteers. 3. Reactivity - Responding in a way b/c they're being observed. 4. Mult Treat Int - When mult IV's effects DV. Needs balancing. |
|
|
Between-Group (S's) Designs
|
The effects of diff levels of an IV are assessed by admin each level to a diff group of S's and then comparing the status on the DV.
|
|
|
Factoral Design
|
When a B-G study includes 2 or more IV and gives more thorough infor about the rel and a main effect.
|
|
|
Main Effect
|
Effect of 1 IV on the DV, disregarding the effects of all other IV. When the marginal means show differences.
|
Self-control example
|
|
Interaction
|
When the effects of an IV differ at different levels of another IV (crossing lines). Requires 2 IV's.
|
Self-control example
|
|
Within-S's Designs (Repeated Measures)
|
All levels of the IV are administered sequentially to all S's. Can include only 2 levels of a IV or can be expanded to include 3 or more levels of a single IV or two or more IV's.
|
|
|
Single-group time series deisgn
|
Type of W/in-S's design. Assess one group sequentially before and after treatment. Threatened by history.
|
|
|
Another type of W/in S's
|
Two or more levels of the IV are applied sequentially to each S and the DV is measured. Ex: Low dose to group and then high dose measuring both times with BPRS. Carryover effects can be problematic, but solved with counterbalancing.
|
|
|
Autocorrelation
|
Disadvantage of time series W/in S's designs due to test-retest effects and increased Type I error.
Observations obtained close together in time from the same subjects tend to be highly correlated. This violates the independence of observations assumption made by statistical tests. |
|
|
Mixed Designs
|
Combines B/t S's and W/in S's methods.
|
|
|
Single-S Designs
|
Each single S design includes at least one baseline phase and one treatment phase. And the DV is measured repeatedly at regular intervals throughout the baseline and treatment phases.
|
|
|
AB Design
|
Baseline and Treatment
|
|
|
Reversal Designs (ABA, ABAB, Etc.)
|
Withdrwal designs b/c treatment is taken away. Provides more data to support inferences if treatment works twice. Can be unethical.
|
|
|
Multiple Baseline Design
|
Sequentially applying a treatment condition to diff beh in diff settings to see if it changes DV. Really an AB in diff settings. Used when withdrawing treatmet is unethical.
|
|
|
Descriptive Statistics
|
Describe and summarize the date collected on a variable or the relationship between variables.
|
|
|
Inferential Statistics
|
Answer the question: can the data be generalized to the gen. pop.?
|
|
|
Continuous Variable
|
Infinite # of values. Ex: Time
|
|
|
Discrete Variable
|
Finite # of values.
|
|
|
Nominal Scale
|
Divides variables into unordered categories. Ex: Male of Female, Eye color, DSM diagnosis, Religion, Political affiliation. Weakness: Only frequencies can be obtained.
|
|
|
Ordinal Scale
|
Places information into "order." Ex: Ranks and Likert-scales. Weakness: Does not tell how much difference b/t scores.
|
|
|
Interval Scale
|
Order and equal intervals b/t successive data pts. Ex: Standard scores on IQ and Temp. No absolute 0.
|
|
|
Ratio Scale
|
Order, equal intervals, and an absolute 0. 0 is the complete absence of the characteristic. Ex: # of calories, # of correct items on a test, & reaction time in sec.
|
|
|
Kurtosis
|
Height or flatness of a distribution.
|
|
|
Leptokurtic
|
"peaked" distribution.
|
|
|
Platykurtic
|
Flat distribution
|
|
|
Mesokurtic
|
A normal curve
|
|
|
Skewed distribution
|
More than half of the observations fall on one side.
|
|
|
Positively Skewed
|
Most scores are low (negateve end) and the positive tail is extended.
|
|
|
Negatively Skewed
|
Most scores are high and the negative tail is extended.
|
|
|
Mode
|
Most frequent score in a set of data.
|
|
|
Median
|
The score that divides the data.
|
|
|
Mean
|
Average
|
|
|
Scales of measurement
|
Nominal-Mode
Ordinal-Mode or Median Interval-Mode, Median, or Mean Ratio-Mode, Median, or Mean |
|
|
Median is used when...
|
the distribution is skewed, b/c the mean is sensitive to all scores (i.e., pull from outliers).
|
|
|
Normal distribution stats
|
68% = 1 SD
95% = 2 SD 99% = 3 SD |
|
|
Range
|
Simplest measure of variability which is calculated by sub the lowest score from the highest score in the distribution.
|
|
|
SD
|
Square root of the variance.
|
|
|
When constants are added or subtracted...
|
the measures of central tendency stay the same.
|
|
|
When scores are multiplied or divided...
|
the measures of central tendency all change.
|
|
|
Inferential Statistics
|
Tells if the obtained sample values can be generalized to the pop w/ confidence.
|
|
|
Population Paramaters
mu sigma sigma squared |
Sample Statistics
M or X S or SD S^2 or V |
|
|
Sampling Distribution
|
Allows a researcher to determine the probability that a sample having a particular mean or other value could have been drawn from a pop with a known parameter.
|
|
|
Sampling Distribution of the Mean
|
Taking several means and finding a normal curve.
|
|
|
Central Limit Theorem (CLT)
|
1. Regardless of the shape of the distribution, as the sample size increases, the sampling distribution approaches a normal distribution.
2. The M of the sampling dist is equal to the pop M. 3. The SD of the sampling dist is = sigma/sq root of N (SEM). |
|
|
The larger the pop SD and the ________ the sample size, the ______ the SEM.
|
smaller, larger
|
|
|
The smaller the pop SD and the ______ the sample size, the ________ the SE M.
|
larger, smaller
|
|
|
Rejection Region
|
Region of unlikely values or your H was right rather than the null.
|
|
|
Retention Region
|
Region of likely values or your H was wrong...keep the Ho.
|
|
|
Type I Error (experiment-wise error rate)
|
False positive. When you reject a true null. Directly related to the size of alpha. As alpha increases, your probability of making a Type I error increases.
|
|
|
Type II Error
|
False negative. When you retain a false null. The probability of making a Type II error is = Beta. When Beta is low, the sample is small, and when the IV is not sufficient, then a Type II error is more likely.
|
|
|
Statistical Power
|
When a statistical test enables an experimenter to reject a false null.
|
|
|
Ways to Increase Power
|
1. Increase Alpha
2. Increase Sample 3. Max IV 4. Min Error 5. One tailed-test 6. Parametric test |
|
|
Parametric Test
|
Are used to evaluate hyp about pop means, variances, or other parameters.
Interval or Ratio scale. Assumptions: 1)Normal dist. & 2)Homoscedasticity (normal variances). Ex: T-test, ANOVA, ANCOVA, MANOVA. |
|
|
Nonparametric Test
|
Used to analyze data collected on variables on a nominal or ordinal scale or when the assumptions of a Parametric test are not met.
|
|
|
Degrees of Freedom
|
N-1
|
|
|
Chi-Square Test (Singel or 1 var & Multiple or 2+ var)
|
Used to analyize the frequency of observations in each category of a NOMINAL VARIABLE. Frequency cannot be less than 5 and obs must be independent.
|
|
|
Single-Sample x^2 Test (Goodess-of-fit)
|
1 var (NOM)
df = c-1 |
|
|
Multiple-Sample x^2 Test
(chi-square test for contingency tables) |
2+ var (NOM)
df = (c-1)(r-1) |
|
|
Mann-Whitney U Test
|
One IV: 2 Ind groups
One DV: Rank order data (ORD) Stasitc: U ALT: T-test for ind samples |
|
|
Wilcoxon Matched-Pairs Signed-Ranks Test
|
One IV: 2 corr groups
One DV: Rank order data (ORD) Stistic: T ALT: T-test for corr samples |
|
|
Kruskal-Wallis Test
|
One IV: 2 or more ind groups
One DV: Rank order data (ORD) Stistic: H ALT: one-way ANOVA |
|
|
T-test for a Single Sample
|
One IV: Single group
One DV: Int or ratio Statistic: T df = n-1 |
|
|
T-test for Ind Samples
|
One IV: 2 Ind groups
One DV: Int or ratio Statistic: T df = n-2 |
|
|
T-test for Correlated Samples
|
One IV: 2 corr groups
One DV: Int or ratio Stistic: T df = n-1 |
|
|
ANOVA
|
Uses to compare 2 or more means and helps control for Type I error.
|
|
|
One-Way ANOVA
|
One IV: 2 or more ind groups
One DV: Int or Ratio Stastic: F df = (c-1)(n-c), where C=levels in IV and N=# of sub |
|
|
F-ratio
|
Mean sq within/mean sq between
(treatment+error)/error When null is true, MSB & MSW are similar. When null is false, MSB is larger than MSW. |
|
|
Factorial (Two-Way when 2 IV's) ANOVA
|
2+ IV's: Indep groups
One DV: Int or ratio Stistic: F |
|
|
Randomized Block Factorial ANOVA
|
When "blocking" is used to control extraneous variables. Treats the ext var as a IV which reduces w/in group variability and increases power.
|
|
|
ANCOVA
|
Combines ANOVA with regression and seperates ext var in the DV.
|
|
|
Repeated Measures ANOVA
|
When using W/in subj designs when diff levels of the IV are admin sequentially.
|
|
|
Mixed (Split-Plot) ANOVA
|
For mixed designs and one IV is B/t groups and one IV is W/in groups.
|
|
|
Trend Analysis
|
Type of analysis of variance used to assess linear and nonlinear trends when the IV is quantitative.
|
|
|
Multivariate ANOVA (MANOVA)
|
1+ IV's and 2+ DV's.
|
|
|
Bivariate Correlation
|
Used to summarize the degree of association b/t two variables. Ex: Scatterplot or Correlation coefficient
|
|
|
Point Biserial Corr
|
True dicotomy such as sex (m/f) and Int or Ratio variable.
|
|
|
Biserial Corr
|
Artificial dicotomy such as climate comfort (fav/unfav) and Int or Ratior variable.
|
|
|
Eta (Used to assess nonlinear relationships)
|
Int/Rat and Int/Rat
|
|
|
Correlation assumptions
|
1. Linearity
2. Unrestricted Range 3. Homoscendasticity (range of x scores is similar to the range of y scores) |
|
|
Shared variability
|
Squared corr coef which represents degree of association.
|
|
|
Regression Analysis
|
Prediction of x and y.
|
|
|
Multiple Regression
|
Multivariate technique for 2+ continuous or discrete predictors.
|
|
|
Multicollinearity
|
High corr b/t predictors
|
|
|
Types of multiple regression
|
1. Simultaneous (simple)
2. Stepwise (step-up and step-down) |
|
|
Cross Validation
|
Trying out a multiple correlation and multiple regression equation on another sample causing the corr coef to "shrink" and decrease the predictive value of the regression equation.
|
|
|
Canonical Correlation
|
Extension of multiple regression that is used when 2+ continous predictors are to be ued to predict status on 2+ continuous criteria.
|
|
|
Discriminant Function Analysis
|
Appropriate technique when 2+ cont predictors will be used to predict or estimae a person's status on a single discrete (nominal) criterion.
|
|
|
Logistic Regression
|
Same as the discriminant analysis, but assumes a non linear relationship.
|
|
|
Path analysis
|
Extension of multiple regression and translates theory about casual relationships into a path diagram.
|
|
|
LISREL (linear structural relations analysis)
|
Used when a casual model includes recursive (one-way) and non-recursive (two-way paths).
|
|
|
Reliability Coefficient
|
In most cases, you would square the correlation coefficient to obtain the answer to this question. However, the reliability coefficient is an exception to this rule: it is never squared. Instead, it is interpreted directly. This means that the value of the reliability coefficient itself indicates the proportion of variance in a test that reflects true variance.
|
|