Stats - Selecting Stats Tests Flashcards by Hayley Mueller

Best measure for central tendency with similar/close scores or values vs best measure for extreme scores or values

Mean is better with similar/close values
Median is better when there are extreme scores/values (i.e., $10,000,000 homes) or substantial percentages of maximum scores

How well did you know this?

Not at all

Perfectly

Idiographic

single subject design/approach

How well did you know this?

Not at all

Perfectly

Nomothetic

group approaches

How well did you know this?

Not at all

Perfectly

What’s Normative data

data that can be compared both within & across subjects

How well did you know this?

Not at all

Perfectly

Ipsative data

results from a forced-choice format (can only describe relative strengths/interests within a subject & not used for comparison across subjects)

How well did you know this?

Not at all

Perfectly

When conducting a ABAB design you should be concerned with:

failure of DV to return to baseline

How well did you know this?

Not at all

Perfectly

What statistical test has nominal data collected for one independent variable?

Chi-Square

How well did you know this?

Not at all

Perfectly

What statistical test has nominal data collected for two independent variables?

Multiple Sample Chi-Square

How well did you know this?

Not at all

Perfectly

What statistical test has interval or ratio data collected for one group of subjects?

T-Test for Single Sample

How well did you know this?

Not at all

Perfectly

What statistical test has interval or ratio data collected for two correlated groups of subjects?

T-Test for Matched or Correlated Samples

How well did you know this?

Not at all

Perfectly

What statistical test has interval or ratio data collected for two independent groups of subjects?

T-Test for Independent Samples

How well did you know this?

Not at all

Perfectly

What statistical test has interval or ratio data collected for more than two groups of subjects are compared to one independent variable

One-Way ANOVA

How well did you know this?

Not at all

Perfectly

What concept refers to the number of possible variations in outcome that can be obtained

Degrees of freedom

How well did you know this?

Not at all

Perfectly

What’s the formula to measure the degrees of freedom in a single sample chi-square?

df = # columns - 1

How well did you know this?

Not at all

Perfectly

What’s the formula to measure the degrees of freedom for multiple sample chi-square?

df = (# rows - 1) x (# columns - 1)

How well did you know this?

Not at all

Perfectly

What formula measures the degrees of freedom for the T-Test for Single Subjects?

df = N (# of subjects) - 1

How well did you know this?

Not at all

Perfectly

What formula measures the degrees of freedom for the T-Test for Matched or Correlated Samples?

df = # pairs -1

How well did you know this?

Not at all

Perfectly

With One-Way ANOVA, what are the 3 different ways of calculating degrees of freedom?
1. total degrees of freedom
2. degrees of freedom between groups
3. degrees of freedom within groups

df total = N-1
df between groups = # groups - 1
df within groups = df total - df between groups

How well did you know this?

Not at all

Perfectly

What’s the formula to calculate the expected frequency in a Chi-Square?

expected frequency = N (sample size) / total # of cells
ex: survey 200 people to see which genders (male or female = 2) vote for Democrat or republican (= 2)
200 / (2 x 2) = 200 / 4 = 50 (expected frequency)

How well did you know this?

Not at all

Perfectly

What is the formula to calculate the expected frequency of a Chi-Square when data is given for each cell?

(Sum of Row x Sum of Column) / N (sample size)

How well did you know this?

Not at all

Perfectly

What’s the risk of running multiple t-tests instead of one-way ANOVA? *will be on test

increases the probability of making a type 1 error

How well did you know this?

Not at all

Perfectly

What’s the possibility of making a Type 1 error if alpha = 0.05 and you run 6 t-tests?

6 (t-tests) x 0.05 = 0.30

How well did you know this?

Not at all

Perfectly

What does it mean if the F-Ratio equals or approximates 1.0 vs if it’s above 2.0, what do they mean with significance?

1.0 = there is no significance
2.0 = there is significance (one group did better than the other group/stronger your results are)

How well did you know this?

Not at all

Perfectly

When the ANOVA is found to be significant, with there being differences between groups, it’s not clear which one of the groups is significantly different from which others; what type of testing is then required?

Post-Hoc Testing

How well did you know this?

Not at all

Perfectly

Which one of the post-hoc tests (Scheffe & Tukey vs Fisher's LSD) provides the least protection from a Type II (2) error, and which offers the most protection?

Least protection: Scheffe & Tukey Most protection: Fisher's LSD

Which one of the post-hoc tests (Scheffe & Tukey vs Fisher's LSD) provides the least protection from a Type I (1) error, and which offers the most protection?

Least protection: Fisher's LSD Most protection: Scheffe & Tukey

What's the downside of Scheffe & Tukey protecting you from Type 1 Errors?

The chance of a Type II error increases

What test do you use when groups are being compared on 2 independent variables (e.g., sex & treatment)

Two-way ANOVA

What test has the advantage of permitting analysis of each independent variable's main effects (significance) but also allows analysis of interaction effects?

Two-way ANOVA

When running a two-way ANOVA, why are 3 F-ratios obtained?

one for each of the 2 independent variables and then another for the interaction effect

How do you know if there was a main effect of a two-way ANOVA when there's a table?

Add up the totals of each column and see if they are the same or different numbers when comparing the totals. If they are different, then that means there was a main effect; if they are the same, there was no main effect

Easy way to see if there were interaction effects with a two-way ANOVA table

Add up two squares that are diagonal from one another, then add up two other squares diagonal from each other; if the sum of each pair of diagonals is different, you can conclude there are interaction effects

When do you use trend analysis as a follow-up to ANOVA, with linear or non-linear data?

To analyze non-linear data

Statistics that depict relationships between variables (X & Y) are termed ________, while statistics that predict are termed _________ or _________

relationships = correlations predict = regressions or analyses

What are the two broad categories of tests of relationship & prediction (hint - variate) and how many variables (X &Y) are in each test

Bivariate: 2 variables (1 X & 1 Y) Multi-variate: involves several Xs and one or more Ys

Which variable (X & Y) is the predictor while the other is the criterion

X (Predictor): This is the independent variable that you believe influences or predicts changes in the other variable Y (Criterion): This is the dependent variable, the one that is being predicted or explained by the predictor (the outcome) e.g., studying the relationship between hours of study (X) and exam score (Y), you might predict the exam score (Y) based on the number of study hours (X)

What does a correlation coefficient test do?

Sees whether there's a relationship between X & Y

A statistical test with the word regression or analysis means we are looking for...

a prediction/to predict something

With a correlation coefficient, what does the negative - or the positive + tell us about the relationship between X & Y?

The direction

With a correlation coefficient, what does the numerical value (between -1.0 & +1.0) tell us about the relationship between X & Y?

the strength

With a correlation coefficient, what does a numerical value of 0 mean vs close value to either +1 or -1

0 = no relationship between the variables close to +1 or -1 = perfect relationship between the variables

What's it called when you square the correlation coefficient?

Coefficient of determination

What does the coefficient of determination represent? (e.g., the correlation between education & income = 0.50)

The amount of variability in Y that is shared with explained by or accounted for by X e.g., - 0.50 (correlation coefficient squared = 0.25 (correlation of determination) - 25% of the variability in income is explained by education - this leaves 75% of the variability in income to be accounted for by other factors like motivation, connections, luck, etc.)

1. How is a simple linear regression (prediction) equation derived in a scatter plot? 2. What does the computer use to calculate it?

1. the line of best fit 2. the least squares criterion

The regression equation is Y= a + bX, what is the "a" and the "b"

a = constant/intercept which determines the height of the line b = slope which determines the steepness of the line

What are the 3 assumptions of a bivariate correlation, and what are they?

1. a linear relationship between X & Y (you can put a line through the dots) 2. homoscedasticity (similar spread of scores) 3. unrestricted range of score (scores must go all the way from low to high)

What calculates the correlation between X & Y when there's a curvilinear relationship?

Eta (old aunt Eta has a curved back)

What type of data does X & Y need to be for a Pearson r statistical test?

Interval/Ratio Data

What type of data does X & Y need to be for Spearman's Rho or Kendall's Tau statistical tests?

Ordinal or rank-ordered

What two tests do you use if one of your variables is nominal (interval/ratio) and the other has a dichotomy (like gender)?

a Point-Biserial (true dichotomy) or a Bi-Serial (artificial dichotomy)

You're correlating a continuous variable and one dichotomous; which 2 statistical tests do you use?

a Point-Biserial (true dichotomy) or a Bi-Serial (artificial dichotomy)

Trick to remembering that Bi-Serial goes with artificial dichotomy, vs Point-Biserial going with true dichotomy

Commercials saying "Buy cereal" and cereals are artificial food/flavours/sweeteners

Which test (Phi vs Tetrachoric) goes with either two variables with true dichotomies or two with artificial dichotomies?

true dichotomies: Phi artificial dichotomies: Tetrachoric

Correlation concept to describe the most basic correlation with no extraneous variables affecting the relationship

Zero Order Correlation

Correlation concept to describe when you examine the relationship between X & Y with the third variable removed that may be influencing or confounding X & Y

Partial or First-Order Correlation

The correlation concept that describes when you examine the relationship between X & Y with the third variable removed from only one of the original variables

Part (Semipartial) Correlation

What is a variable that influences the strength between the predictor and criterion?

Moderator variable e.g., age would be the moderator variable with old age & smoking = stronger correlation vs young age & smoking = weaker correlation

What is the variable that explains why there's a relationship between the predictor and the criterion?

Mediator e.g., if education explains the link between SES and smoking, when you control the mediator (education), then the relationship of X & Y becomes almost zero

What test involves the correlation between two or more IVs (Xs) and one DV (Y), where Y is always interval or ratio data and at least 1 X is also interval or ratio data

Multiple R (or multiple correlation)

How do we get the coefficient of multiple determination?

squaring R (multiple R score)

What's the concept/equation prediction equation to make a prediction when we have 2 or more Xs and one criterion?

Multiple regression equation Y = a + b1X1 + b2X2 +b3X3

Multicollinearity is a problem associated with what multivariate test and what does it mean

- associated with multiple regression - it happens when the predictors are highly correlated with one another

What is the ideal outcome with correlations in a multi-regression: 1. between the predictors (Xs) 2. between each predictor & the criterion (Y)

1. low correlation between predictors 2. moderate to high/strong correlation between each predictor and the criterion

Is multiple regression a compensatory or non-compensatory technique?

Compensatory technique

Difference between the 2 subtypes of multiple regression, stepwise regression & hierarchical regression

1. Stepwise regression: the computer decides which variable to put in first, second, third, etc., based on the strength of the correlation 2. Hierarchical regression: the researcher controls the analysis, adding in the order based on the theory the researcher is using

Canonical R is the relationship between how many Xs and how many Ys?

two or more IV's (Xs) and two or more DVs (Ys)

The discriminant function analysis is used with how many Xs and how many Ys, and what does the Y variable have to be?

two or more predictors (Xs) one criterion (Y) - a nominal (categorical)

What type of analysis is used to predict nominal/categorical criterion (Y) based on nominal/categorical predictors (Xs)? e.g., type of graduate school (X1) & sex (X2) predicting the likelihood of passing or failing the EPPP

Loglinear or logit analysis

Path analysis and structural equation modelling are ways, using correlational techniques, are ways to determine _______ relationships between variables

Causal relationships

Factor analysis and cluster analysis are tests of ________

Structure

What are tests of structure for?

when a researcher is interested in discovering which variables in the set best fit together or form coherent subsets

What test is used to reduce a large number of variables into a smaller number of factors?

Factor analysis

What does an Eigenvalue (characteristic root) tell you in a factor analysis? and if it's less than 1.0 what does that mean?

Eigenvalue tells you the strength of the factor less than 1.0 is considered NOT significant

The factor loadings of a factor analysis are interpreted and to make the factors more distinct and interpretable, what is used?

Factor rotation

What two types of rotations are used?

Orthogonal and oblique

With orthogonal vs oblique rotations in factor analysis, which one relates in factors with correlation or no correlation

Orthogonal = no correlation Oblique = correlationn

Difference between principal component analysis vs principle factor analysis

Principal component analysis: the researcher has no prior hypotheses Principle factor analysis: there's an underlying theoretical basis before the analysis

What analysis is involved in gathering data on various dependent variables and statistically looking for naturally occurring subgroups in the data?

Cluster Analysis

What test has the main assumption that there's independence of observations?

Chi-Square Test

What type of tests have these three assumptions: data needs to be interval/ratio, homoscedasticity must be present, and the data must be normally distributed?

Parametric tests

Is a Chi-Square test parametric or non-parametric?

non-parametric

What helps researchers make inferences about causation & includes LISREL as a type of it?

Structural Equation Modelling

What can test out many causal pathways involving multiple predictors and criterion variables?

Structural Equation Modelling

Which one controls carryover effects when repeated measures are used vs which one controls for the effects for testing (Solomon four group vs the latin square)

Solomon four group = controls for the effects of testing Latin-square: controls carryover effects when repeated measures are used

A researcher would use LISREL, a form of structural equation modelling, in an attempt to: 1. test a causal model of relationships among variables 2. derives a causal model of relationships among variable

TEST a causal model of relationships among variables