Upgrade to remove ads
Only $35.99/year

Psyc 4418- Final

Terms in this set (284)

_____ is observed when an IV is associated with Multiple R of .95 in multiple regression which includes all other IV's as predictors, making the IV redundant as the IV is over-correlated.
Multi-collinearity
_______ represents standardized beta and is used to get effect size in multiple regressions.
Semi-partial correlations
In curvilinear multiple regression, a negative semi partial correlation implies
an inverted u shape or negative threshold
The value of DV when IV is 0
Intercept
In Regression,_____ represents the standard deviation of residuals (error of prediction)
Standard Error of Estimate
In regression, tolerance should not be less than ____ and VIF should not be greater than ___.
.10, 10
_____ is the percentage of variance in the IV that is unique, while ______ is the degrees to which beta weight Standard Error is larger than it would be without IV correlations (tests of multicollinearity.
Tolerance, VIF
Should we worry about multicollinearity in moderation/curvilinear regression? Why?
No. It is an illusion.
In lecturers opinion, what is the best way to report effect size in multiple/curvilinear regression?
use semi-partial correlations
Can both a linear and curvilinear effect be interpreted from data?
No.
In significant curvilinear effects, the lecturer states interaction is present because...
The magnitude and direction of x on y varies as a function of x
According to the lecturer, is there a difference between moderation and interaction? Why?
No. Both mean the magnitude of effect of IV on DV varies as a function of another IV.
What are the three variables in moderation?
X, M, and Y
o-->x <---->
o---->Y
o-->M<--->

This path analysis displays....
moderation
Testing interaction in context of multiple regression is known as...
moderation analyses
In logistic Regression we can obtain both ____ and ____ signifying a 1 standard deviation increase in X is associated with beta difference in Y. But we can not determine ______ without further analysis (e.g bootstrapping)
standardised and unstandardised beta weights, statistical significance
We can only run multiple regression on a nominal DV with ____ point scale (or more), but if less the main concern is _______ _______.
5, accuracy of standard error
Technically, you can still use dichotomous DV's in multiple regression if you use _______.
bootstrapping
Psychological Dimensions are discerned through _______
Data Reduction Analysis
Data Reduction is interested in uncovering the _____ and _______ of dimensions that differentiate individuals
number, nature
The most popular type of data reduction analysis is ______.
Factor analysis.
In exploratory factor analysis you don't have an idea of the _____ or the___ of dimensions in data.
number, nature
An ____ is a variable for which each case has an individual score. This includes:
Observed variable; items and subscales/composite scores
Theoretically, in the context of factor analysis, patterns in scores are (assumed to be) cause by _____ which causes the scores to _____.
dimensions, correlate
In _____ _______, an alternative data reduction technique, no assumption is made about what has caused the observed variables to inter-correlate.
Component analysis
What is the path analysis for Moderation?
What is the path analysis for Classical Suppression?
oj
What is the path analysis for mediation?
What is the path analysis for Component Analysis?
What is the path analysis for Factor Analysis?
What is the path analysis for Coopperative Suppression?
What is the path analysis model for bivariate regression?
In component analysis there is ____ with a component, but no ____.
correlation, causality
In Factor analysis, a factor has an effect on observed scores AND there is a _____ associated with the analysis.
error term.
The only way an error term can appear is if _____ because there is ____.
an observed variable is being hypothesized, a residual chance the hypothesis is wrong.
What are the 7 steps to data reduction (factor analysis)?
1.Determine whether the data are appropriate for reduction
2.Decide the number of dimensions to extract
3.Decide estimation method to generate solution
4.Decide how to rotate the solution
5.Interpret the dimensions by examining salient loadings
6.Examine correlations between dimensions
7. Consider possibility of running second-order data-reduction analysis.
The fundamental basis of data reduction technique is the ____ associated with the observed variables. If variables are not ____ you cannot perform data reduction.
correlation matrix, correlated
What 2 tests help us identify if we should proceed with data reduction?
Bartletts Test of Sphericity and KMO
In factor analysis, ___ tests whether at least one correlation is significant.
Bartlett's test of sphericity.
Bartlett's test of sphericity is heavily influenced by ____ and can be excessively ____. It is not recommended by the lecturer.
sample size, powerful
In factor analysis,_____ tests the hypothesis that the correlation is an identity matrix (matrix of zeros) within sampling fluctuations.
Bartlett's test of sphericity.
mediation, indirect effects, and suppression all fall into category off ____.
process analysis
can curvilinear and moderator effects be confounded? if so, how can this happen?
yes, if the IVs are highly correlated.
A ____ effect is the opposite of ____.
suppressor, mediation
I'm suppressor effects, instead of reduction in size of total effect, an affect _____ in size by introduction of another variable.
increases
occurs when there is a small amount of shared variance between IV and DV, a moderate amount of variance between another variable and the IV, and little to no variance between the third variable and the DV.
classical suppression
____ may be argued to be tested by comparing the difference between the squared zero-order correlation between x and y and the squared semi-partial correlation between x and y when controlling for the effects of a third variable on x.
classical suppression
occurs when x is weakly interrelated with y, a third variable is weakly interrelated with y, and the third variable is more strongly interrelated with x.
cooperative suppression
In _____ both x and s are mutually suppressed
cooperative suppression
when cooperative suppression is observed both ____ and ____ are included in multiple regression to y, their beta-weights are observed to____ in magnitude in comparison to zero-order correlations to y.
x and s, increases
in cooperative suppression you would consult ____ while in classical suppression you would consult _____.
beta weights, semi-partial correlations.
according to paunomens and Lebel, a ____ percent increase is moderate and a _____ percent increase is considered _____.
1, 2
the probability of rejecting the null when the null is true.
Power
In _____ participants are tested only once on 1 or more dependant variables while in _____ participants are tested more than once on 1 or more DVs
between subjects design, within subjects design
In within Subject dressings the shared variance between the conditions ____ the standard error of the difference between the means
reduces
what are the two ways to create adjusted standard deviations fir calculating confidence intervals in within subjects designs.
individual and pooled
1.comducf a 1-way within subjects ANOVA
2. Square route the mean square error
3. transform raw scores to z-scores
4. transform z scores to adjusted scores with raw score means and adjusted SDs.
these are the steps to obtain:
pooled adjusted standard deviations
What is the formula for pooled adjusted SD?
square route of the mean square error
In a 2x2 Factorial ANOVA the assumption of sphericity is ____.
satisfied
In ANOVAs with 3 or more levels, what test can we use if we fail to satisfy the assumption of sphericity?
Huynh-feldt
How can one interpret main effects in the presence of of interaction? (within subjects design)
By conducting simple main effect analyses/paired sample t-tests
If the simple main effect analyses are _____ and in the _____ direction across the levels of an IV, the main effect can be interpreted in the presence of interaction.
statistically significant, same
Ideally, we would be able to use _____ term for each of the simple main effect analyses to uncover a main effect in the presence of interaction. However, SPSS does not allow such an analysis.
pooled error term
What is the quick and dirty approach to uncover simple main effects in the presence of interaction?
conducting a series of paired sample t-tests
What is the pooled error term approach to uncovering simple main effects in the presence of interaction? Note: this must be done by hand
convert MSe into a pooled standard error of the difference between related means.
It is argued that ordinary least squares multiple regression cannot handle ____ scored DV's because the ____ and the _____ standard errors cannot be trusted to be accurate.
dichotomously, R square and beta weight
Prior to logistic regression researchers used _____ which is very similar to ______.
Discriminant function analysis, MANOVA
A key limitation of DFA is...
equality of covariance matrices across DV groups
In ____ mathematics involve ordinary least squares estimation while in logistic regress the fundamental mathematical basis is____.
multiple regression, the natural logarithm of an odds ratio.
Chi square analysis is also known as
Contingency table analysis
In contingency tale analysis, pearson chi square tests____.
The difference between 2 percentages.
In contingency table analyses, ___ is a correlation on dichotomously scored variables.
Phi
Another method to describe a 2x2 contingency table analyses is known as an _____.
odds/ratio
When there is _____ an odds ratio will be 1.0.
Total absence of an effect.
What form of regression can be used if there is a dichotomously scored DV?
logistic regression
If an odds ratio is greater than 1 you might say....
(Use example odds ratio = 2.01 boys versus girls, DV recommendation remedial reading).
Boys are 2.01 times more likely to be recommended for remedial reading.
An odds ratio is the _____ of a correlation coefficient in which a score of 1 means a 1 to 1 correlation, but in odds/ration it means a ___ to ___ relationship.
opposite, 1 to 0
In multiple regression we look at DVs that are _____ or ______.
continuous, ordinal scales with 5 or more groups.
The formula for an odds/ratio is OR=______where a,b,c, and d correspond to _____. This formula provides a ratio but not level of significance.
(a/b)/(c/d), frequencies in contingency tables
In odds ratio, a full inter-association box/matrix displays...
correlations between variables
In order for a logistic regression to be significance it needs to....
beat the classification accuracy of a base model in which only knowledge of DV is used to classify cases.
An omnibus test of model coefficients can be viewed as a test of _______ of the ______ _______. This uses ___ ____.
statistical significance, entire model, chi square.
The _____ and ____ are ____ to model R squared in multiple regression.
Cox Snell R square, Nagelkerkie R square, similar but not equivalent
The ____ R squared is always larger than the ____ R squared because it includes an adjustment to allow for an R squared of 1.0, meaning it allows for the possibility of ____ % prediction. (the lecturer recommends reporting both)
Nagelkerkie, Cox and Snell, 100
The ____ test is an alternative test of statistical significance of overall model when at least 1 or more IV is scored on a continuous scale. This test is significant when p is ____ than .05.
Hosmer and Lemeshow, greater
In SPSS logistic regression does not report ______, but these can be calculated with this equation: bSDIVR/SD logity. The last part of this equation is the standard deviation of ___________.
standardised beta weights, predicted log odds.
____ is used to compare which variable is more important to a model when the variable are on different scales by transforming the scores to Z-scores. IF the variables are on the same scale you can use ____.
standardized beta weights, unstandardized beta weights
Semi partial correlations are used to obtain ____.
effect size.
_____ regression is hypothesized through theory and is best seen on a scatter plot.
Curvilinear
_______ represents a 1 standard deviation increase in the value of X is associated with change (positive or negative) in Y
beta weights
A linear effect implies the ___ and ____ of association is the ____ across the IV and DV scores.
direction and strength, same
Pearsons r is a_____ coefficient which can also be called _____.
correlation, zero-order correlation.
A ____ is an association between 2 (or more variables)
correlation
In bivariate regression the beta weight and Model R are ______ because they represent the percentage of variance accounted for by _____.
the same, the whole model
In Bivariate regression, Model R is...
a pearson correlation.
A bend in the line of best fit suggests a _____ association between 2 variables.
curvilinear
What are the 4 types of curvilinear effects?
U shape, inverted U shape, Positive threshold, negative threshold.
According to lecturer, which curvilinear effect is most common in psychology. What economic term does this phenomenon display?
positive threshold, law of diminishing returns
Why is it misleading to report pearson correlation is present across all data in curvilinear effects?
The strength and direction of association varies throughout the curve.
A pearson correlation is the product of ____.
2 Z-scores divided by N-1
A quadratic term implies ____ bend(s) in the line of best fit.
1
In moderation, the ____ effect is added at step 1 and the ____ effect(s) is entered at step 2.
main effects, interaction effect/product term
In moderation ____ represents the percentage of variance accounted for by the product term, while ___ represents the unique variance accounted for by adding the product term.
R square, R-square change
When a zero-order (pearson) correlation and beta (standardised) are similar in regression/moderation analysis, one may assume the Independant variables are ____.
unique
In _____ one splits samples across the moderator variable and plots the effect between the IV and DV with 2-3 different slopes. This is the best way to plot moderator effects.
spotlight analysis
In _____ the IV varies as a function of the DV.
moderation
In moderation, we should consult ________ to evaluate the magnitude of effect of product term on data.
semi-partial correlations
______ decomposes model R square into unique and common effects due to individual predictors and combination of predictors. It is used to evaluate the magnitude of effects in moderation. The lecturer does not support this analysis.
communality analysis
An intercept is the value of ___ when ____ is 0.
Y, X
____ is a series of squares, lines, and circles that represent terms of a regression visually.
path analysis
In____ 2 or more IV's predict a continuously scored DV.
multiple regression.
In multiple regression, the ______ represent the unique contribution of each IV to regression equation. These need to be tested for statistical significance.
beta-weights
A limitation of multiple regression is that it is a _____ not a(n) _____.
prediction, explanation
A ____ variable can help explain why an IV is related to a DV because it accounts for some of the variance between X and Y.
intermediary
A ______ is theorized to be the mechanism by which the IV has influence on the DV.
mediator
In mediation, ____ is the Pearson correlation between X and Y.
Total Effect
In order for mediation to be present, there must be a significant _____.
total effect
The mediator variable is hypothesized to ___ between the IV and the DV.
intervene.
The analyses of mediation decomposes the total effect between X and Y into 2 parts: _____________ and _____________.
direct effect, indirect effect.
_____ is the effect between x and Y that is not explained by the intermediary variable, while the ____ is the effect between x and Y that is explained by the intermediary variable.
Direct effect, indirect effect
In partial mediation the total effect is ____ and numerically ____ than the direct effect.
significant, larger
B1y represents ____ and B1m*B2y represents _____.
direct effect, indirect effect
How can we determine the significance of beta weights for indirect effects?
Sobel test and bootstrapping
A ____ test uses unstandardized beta weights to determine the significance of an indirect effect. This test is strict (lacks power) and assumes a normal distribution of data. (mediation)
Sobel test
_____can be used to calculate standard error from a large series of resampled data.
bootstrapping
By using ____, the results are accurate irrespective of level of non-normality (e.g skewness, kurtosis).
bootstrapping
______ creates an empirical sampling variability distribution from raw data.
bootstrapping
In bootstrapping, an indirect effect is significant when the ____ are in the ____ direction
confidence intervals, same
The key characteristic of ____ is that all direct effects associated with the model have to be tested individualy first and observed to be statistically significant.
Causal steps approach
Baron and Kennedy's causal step approach is inappropriate to use because it does not distinguish between mediation and _____ and fails to recognize that some effects are due to a ______.
total indirect effects, suppressor
_____ occurs when the magnitude of difference between two or more means depends on the level of another IV. Alternatively, this occurs when the magnitude of the difference (Cohens D value) associated with 2 (or more) means is statistically significantly different from Cohens D of other 2 (or more) means.
interaction
In odds/ratio a negative correlation indicates _____ scores on levels of IV associated with ___ in dichotomous DV (where the levels of DV are coded as 0,1)
higher, 0
_____ is the unique association between IV & DV controlling for effects of the other variables.
effect size
The calculation of CI in within subjects designs is complicated because you must take into consideration the __________ between conditions.
shared variance
1.estimat variances associated with the conditions and the covariance
2.subtract the covariance from variance
3. square root adjusted variances
4.transform raw scores to Z-scores
5. transform Z-scores to adjusted scores
These steps are used to create:
Individual adjusted SD
Multiple regress usually uses ____ estimation technique while logistic regression uses a ______ estimation.
ordinary least squares, maximum likelihood
It is often argued that Ordinary least squares estimation cannot handle ____ scored DV's because _____ are inaccurate and you can get misled by ____ (the lecturer does not support this). It is better to use logistic regression.
dichotomously, standard errors, p-value
In logistic regression, the key matrix to look at is the variables in equation matrix. This matrix presents an alternate version of a T test called a _____ statistic. It also presents the ____ which is called 'constant' in SPSS.
Wald, intercept.
In logistic regression, the Exp (B) column provides the ____ odds ratio results meaning it ____ for the effects of other variables.
unique, controls
The ____________ is the best way to discern which predictor is the most important to the equation, but there is no test to discern ______.
Wald statistic, statistically
1. Create log odds variable: Intercept+ (IVb1)+(IVb2)...
2.Convert to predicted odds: Exp(log_odds)
3. Compute probabilities: predicted odds/(1+predicted_odds)
These are the steps for transforming ________ into _________.
odds ratio, predicted probabilities
Predicted probabilities less than ____ for a dichotomous variable scored 0,1 indicated absence of whatever 1 is measuring, in this case recommendation for remedial reading.
.5
In logistic regression, group membership can by determined by ______ predicted probability into values of 0 and 1. This predicts future scores.
rounding up or down.
_______ is used in moderation to reduce correlation between the product terms and the IVs. The lecturer does/does not recommend this.
mean centering
_____ is the variation of presentation of conditions across within subjects designs. It is used to reduce ordering and time related effects.
Counterbalancing
When no rotation is specified you get _________. When an oblique rotation is specified, two additional factor loading matrices are reported: __________ and _____________.
Factor matrix, pattern matrix, structure matrix
In _____ IV's must be scored on a nominal scale while in ____ the IV can be measured on a nominal or continuous scale.
ANOVA, multiple regression
In moderation, Model R represents the percentage of variance that was accounted for by the ________.
product term
What analysis do you use to examine moderation?
hierarchical multiple regression
_____ occur when there is no total effect between X and Y but there is a significant indirect effect from x to m to y.
Total indirect effect
______ is observed when the direct effect, indirect effect and total effect are significant.
Partial mediation
In multiple regression, IV's can be scored on _____ scales, but the DV must be scored on a ______scale.
continuous or nominal, continuous
A model developed to predict a continuously scored DV from 2 or more IVs is ____.
multiple regression
____ is observed when the direct effect is non-significant and the total effect is significant.
Total mediation
In preacher and Hayes syntax output, the "bootstrap results for indirect effect" uses what to analyze significant while the? What do you use to determine significance?
bootstrapping, confidence intervals
A more inclusive approach than causal steps, known as ______, does not require testing any direct effects first for statistical significance.
process analysis
The fundamental basis of standard error is...
variability in point-estimate estimation across random resamples drawn from the population
In odds ratio, the ____ the deviation from 1, the larger the effect between 2 variables.
larger
In odds ratio it is easier to interpret odds ratios _______ than 1.0
greater
What is the best way to present an interaction in ANOVA
plot the means.
If an odds ratio is less than 1 you might say option 1 is ________ percent less likely to be.... (only answer the blank)
1-number
Pattern matrix represents the _______________of each variable onto each other. Thus the pattern coefficients represent _______, just like beta weights in multiple regression.
standardized regression weights, unique effects
The ________ _________ and _________ _________ are rotated while the ________ ___________ is not.
pattern matrix, structure matrix, factor matrix
Pattern Matrix coeeficiants should be at least _____ or larger in magnitude. If over _____ is considered very larger. Most fall between _________ and ____________.
|.20|, |.80|, |.30| and |.70|
The sum of squared loading only equals communality in the _______ matrix.
factor
In factor analysis, the __________ coefficients represent the correlation between a variable and a factor.
structure matrix
Because the structure matrix uses correlations, structure coefficients (loadings) tend to be __________ than pattern coefficients.
larger
In the structure matrix if 2 variables are correlated with a factor, you can also assume...
they are correlated with eachother
In factor analysis there are ____ types of correlations. They are:
3 types
observed correlations, reproduces correlations, residual correlations
Observed correlations are found in the ___________ and represent regular correlations between variables. These indicate whether it is appropriate to perform data reduction.
correlation matrix
Reproduced correlations are found in the reproduced correlation matrix and represent _________ between observed variables across factors.
sum of cross-products
A good factor analytic solution will be associated with reproduced correlations that are ____________ the corresponding observed correlations.
similar to
A ______ between the 2 types of correlations (observed and reproduced) would suggest that a very large percent of the common variance between observed variables has been accounted for by the factor solution.
close correspondence
_____ represents the difference between observed correlations and reproduced correlations.
residual correlations
A residual correlation greater than _____ is suggestive that non-negligible amount of common variance remained to be determined by the factor analysis.
|.10|
The lecturer states you should conduct a second-order factor analysis if...
the factors are inter-correlated
There are ____ types of factor scores which can be developed in SPSS. These are:
3
regression, bartlett, anderson-rubin
Regression factor scores are based on _____ matrix coefficients and includes the ______.
structure, inter-factor correlations
____ factor scores constrain the factor scores to be uncorrelated with other factors while _________ factor scores constrain the factor scores to be uncorrelated with other factors AND with eachother.
bartlett, anderson-rubin
For factor scores, the ____ method is best because it provides factor scores that are ____ with the factor of interest. But this method can also share factor scores with other dimensions. This can be resolved through ____________.
regression, most correlated, multiple regression
A scree plot is a scatter plot of eiganvalues on the ____ axis and the __________ on the ______ axis.
y, rank-ordered eiganvalues, x
Scree represents___________ _____________.
Random variance
In SPSS, the scree plot is based on ______ ______ eiganvalues.
Component analysis
The scree plot leads to both ___________ and __________ of dimensions. This is mainly because scree analysis is as much an art as a science.
over extraction, under extraction
In parallel analysis, the number of dimensions to extract is equal to the number of ___________ above the ________.
sample data eigenvalues, scree
____________ simulates a larger number (minimum 1000) of random (rubble) data sets based on the characteristics of the original sample data.
parallel analysis.
In parallel analysis, you can use both ______ and _____ percentiles to calculate random data sets, but ______ eiganvalue percentiles is more rigorous. One is not necessarily better than the other.
50th, 95th, 95th
Parallel analysis identifies the number of factors correctly only about__________ percent of the time.
75
In SPSS parallel analysis compute kind 2 runs the data as a ___________ approach of components, while 1 equals__________ approach. More research needs to be conducted on which is better.
factor estimation , principle components
________ represents the percentage of common variance between the observed variables that is accounted for by the number of factors extracted from the correlation matrix. This is the best method for determining number of factors to extract.
Goodness of fit index
The formula to calculate GFI is 1- (X squared implied/X squared null) where X squared implied is _____ ____ ______ and X squared null is _______ ____ ______.
GFI chi square, Bartlett's chi square
GFI values less than ____ suggest there may be 1 or more additional dimensions to extract.
.95
The bigger the Bartletts Test of Sphericity Chi square value, the ______ and _____ correlations there are in the correlation matrix.
more, bigger
In maximum likelihood estimation you want GFI P to be ________ because this p-value indicates there is not even _______.
non-significant, 1 correlation left in the residual correlation matrix
SPSS offers ____ options of data reduction estimation methods. The most popular are:
7
principal component, principal axis factoring, maximum likelihood
_________ estimation method is not a factor analysis.
principal components
The key difference between component and factor analysis is that a component analysis includes the __________ associated with each variable while factor analysis includes only the _________ between the observed variables in the analysis.
total variance, common variance
Prinipal Axis Factoring is similar to component analysis except the analysis is conducted on the _________ which replaces the 1.0 on the diagonal of the correlation matrix with ____________ which are less than 1.0.
reduced correlation matrix, communalities
By replacing the diagonal with common variance, PAF ensures a ______________ solution is obtained.
common factor model
MLE is a sophisticated _____________ estimation technique that has been shown to often yield results that best reflect the population.
common factor
MLE is better than PAF when:
1. the factor loading within a factor are ___________ (|.40| to |.70|
2. Factors are ___________
meaningfully unequal, inter-correlated
___________ represents the percentage of variance associated with an indicator that was contributed to the factor analysis. This is represents by the _________ column in SPSS and is symbolized as ________.
communalities, extraction, h squared
In SPSS communalities output ____ represents the amount of shared variance a variable has with other variables in the analysis (similar to R squared) and _______ represents the communality based on the factor analysis.
initial, extraction
If the observed variables are items communality should be greater than ____. IF the observed variables are subscales communality should be greater than ____. This implies #% of item/subscales variance has contributed to the _________.
|.04|, |.09|, factor solution
In the total variance explained output, the initial eigenvalues represent ________ while the extraction sum of squared loadings represent ________ so the initial eigenvalue column should be ignored.
component eigenvalues, extracted dimensions
Any factor that accounts for less than ______ percent of unique common variance should be considered rather small. You may want to remove this from your factor analysis if it will not compromise GFI as it is always better to over extract than under extract.
5
_____ represents the quantitative association between an observed variable and a factor. If greater than _____ this indicates overextraction.
Factor loadings, |1.0|
The factor matrix is like ________ representing the _____ each observed variable is contributing to the factor.
beta weights, unique variance
A factor made up of all positive loadings is a ______ which is defined by all observed variables.
general factor
__________ is achieved when an observed variable has one salient loading (<.\|.50|) on a single factor and non-salient loadings (>|.20|) on the remaining factors.
Simple structure
Low simple structure is when there are _________ and high simple structure is when_____________.
multiple indicators with loadings on multiple factors, each indicator with salient loadings on a single factor
Rotation techniques increase __________ of a factor solution by _________.
simple structure, distributing the variance in a different way.
The two broad types of rotation methods are _____________ and _____________
orthogonal, oblique
__________ rotation methods force factors to be correlated while _______ rotation does not force factors to be correlated, but allows them to be correlated.
Orthogonal, oblique
According to the lecturer, ____ rotation is preferred because _______.
oblique, we expect factors to correlate
__________ is a popular method of orthogonal rotation.
Varimax
There are 2 popular oblique rotation methods: _________ and _________
Promax, direct oblimin
Promax starts of with _____ rotated solution and then rotates toward a target computed as current factor loading taken to a higher power, specified as ________.
varimax, kappa
In Promax, Kappa is best specified between _______ and _______. The larger the kappa specification the larger _______ between factors can be.
2 and 4, correlations
_______ rotation tries to minimize both the squared cross-products of factor loadings and sum of the covariances of the squared cross-products is minimized.
Direct oblimin
In plain english, direct oblimin strikes a balance between _________________ and ________________________
achieving simple structure, producing a factor solution that might be excessively correlated.
In direct oblimin delta needs to be specified. Delta can be as low as _____ and as high as ____. Lower numbers present a(n) ________ and higher a(n) _________ factor solution
-4, .80, orthogonal, oblique
In Bartlett's test, the data is technically okay for reduction if p is _______________. In KMO p must be _______________ for appropriate data reduction.
less than .05, greater than .65
According to the Kaisser Guttman Rule, an eigenvalue greater than 1.0 implies ___________.
a dimension with common variance equal to more than a single observed variable.
The Kaisser Guttman Rule rule leads to ___________ of dimensions, especially when mean correlations are less than _________ in magnitude.
overextraction, |.30|
A __________ Bartlett's test is a strong indicator that data is not appropriate for data reduction.
non-significant
The ______ states the number of components to extract should be equal to or greater than 1.
Kaisser Guttman Rule
The __________ test of sampling adequacy values range of .00 to 1.0 in which a value of .5 implies an average correlation of ____. The test is acceptable for reduction if p is _____.
KMO, nothing, greater than .65
The KMO test is more valid when the case to observed variable ratio is _____ to _____.
5 to 1
What tests can be used to determine number of dimensions to extract in factor analysis?
scree plot evaluation, Kaisser-Guttman Rule, parallel analysis, Goodness of Fit Index
An ______ represents the amount of variance explained by the dimension
eigenvalue
Eigenvalues are not scaled in ____________. Instead they range in value from .00 and _________________.
percentage terms, the number of observed variables included in the analysis
in more precise terms __________ represents the sum of the squared loadings.
eigenvalues
When the scales are meaningful, standardization can ________________ making a variable that shouldn't yield a cluster produce clusters.
obscure differences
There are 2 types of clustering. They are ________ and __________.
hierarchical, non-hierarchical
_________ reduces a large number of variables into a smaller number of dimensions while _______ reduces a large number of cases into a smaller number of groups.
factoring, clustering
Both factoring and clustering perform
dimensionality reduction
Both factor analysis and cluster analysis are multivariate statistical techniques that create
groupings
In ________ the emphasis is upon creation of clusters defined by _______ which load onto those clusters.
cluster analysis, cases
Factor analysis is based on a _______ matrix while cluster analysis is based on a ________ matrix
correlation, proximity
Factor analysis shows the ______ between observed variables while cluster analysis is often based on __________ among cases.
similarity, dissimilarity
There are no factor loadings in __________ analysis.
cluster
1. select variables for analysis
2.select similarity/dissimilarity measure
3. select type of analysis
4. select type of linkage
5. evaluate number of clusters
6. determine natures of clusters

These are the steps to conduct ________.
cluster analysis
Cluster analysis is based on identifying patterns of ____________/__________ across variables that homogenize and separate cases within and between clusters.
similarity/dissimilarity
In cluster analysis, the inclusion of both________ variables and _______ inter correlated variables can cause cluster analysis to be unbalanced.
irrelevant, highly
If you have highly inter-correlated variables, it is probably best to run a _____ instead of a cluster analysis.
factor analysis
In cluster analysis, there are __________ different measures of similarity/dissimilarity. They are:
8
euclidian distance, squared euclidian distance, cosine, pearson correlation, block, chebyshec, minkowski, customized
Cases with _____ tend to be assigned to the same cluster while cases with ________ tend to be assigned to different clusters
high similarity, low similarity
In cluster analysis, we expect that each group is maximally ______________ with respect to 2 or more characteristics. We (also expect) groups are maximally _________ from other groups with respect to 2 or more characteristics.
homogenous, distinct
_____________ and ________ are the most common distance measures for cluster analysis in psychology.
Euclidian, squared euclidian
________ represents the sum of the _____________ difference scores across all of the variables between two cases
euclidian distance, absolute
_____________ represents the sum of the squared difference scores across all of the variables between 2 cases.
squared Euclidian distances
Almost all studies choose a ______ distance measure which is appropriate for both _____ and ___________ scored data.
euclidian distance, binary, continuous
____________ are particularly appropriate for centroid, median, and Ward linkage methods.
Euclidian distance measures.
Euclidian distance may inaccurately measure distance between cases if the variables are _______________. As a result you may have to transform the cases to _____________.
scored on different scales, z-scores
There are ______ approaches to heirarchical regression. They are.
2
agglomerative, divisive
Heirarchical clustering is ___________ in nature and is best to choose when the number of clusters is ___________ and the number of cases is ________.
exploratory, unknown, not very large
Non-heirarchical cluster analysis is best chosen when the number of clusters is ___________. Thus it is less __________ in nature. I is more useful when the number of cases are ______.
known, exploratory, very large.
the agglomerative approach _______ while the divisive approach __________.
combines, divides
____________ is a bottom-up approach in which each case starts off as an independent cluster and the closest cases (K-1) are combined together until they are all clustered into a single group. _________ is a top-down approach which does the opposite.
agglomerative approach, divisive approach
There are ___ types of linkage in SPSS. They are
7
between groups linkage
centroid clustering
median clustering
within-groups linkage
nearest neighbour
farthest neighbour
Ward's method
__________ linkage can only be used if data is on a interval/ratio scale and euclidian distance is used.
Centroid clustering
______ linkage is also known as average linkage. It works well with euclidian distances, is less affected by outliers, and tends to produces clusters with small within cluster variation
between-groups
_________ of linkage is the most popular method which looks at change in total within-cluster variance of clusters when we join them.
Ward's method
Ward's method is easily affected by _________, works best when the number of cases per cluster is equal, and is most similar to non-hierarchical cluster analysis called ________.
outliers,equal, k-means
To determine the number of clusters, create a _______ of the number of staged (n-1)/cases on x-axis and the _______on y-axis
scree plot, agglomerative coefficients
In a cluster analysis scree plot, a substantial _____ in the coefficient suggests a genuine series of clusters
Jump
To understand the nature of clusters, conduct a _________ with the variables as predictors and ____________ as a dichotomous/nominal DV.
multiple regression, group membership
In the agglomeration schedule, coefficients tell us the _________ or _________ as we increase clusters/join things together. Higher coefficients indicate larger ____________ in the solution.
fit or change in fit, error
In an agglomeration schedule __________ represent the amount of distance between the two clusters/cases joined together. The larger the ___________ the greater the benefit there is to separating cases/clusters into smaller clusters.
coefficients, coefficients
_________ is a graphical representation of clustering that shows which cases are aggregated and the merging distance between cases.
dendrogram
____________ is scaled by joining distances of clusters.
dendrogram
To understand the nature of clusters, consult _______ from a multiple regression in which cluster/group membership is the DV.
beta weights
_______ is the most common type of non-hierarchical cluster analysis.
k-means
K-means is not based on a ___________ between clusters
measure of similarity/dissimilarity
In k-means you have to specify the ________.
number of clusters
k-means can handle _______ numbers of cases, the results are __________ by inclusions of irrelevant variables and by the measure of similarity/dissimilarity.
larger, less impacted
People often use _____ to deduce number of clusters to specify in k-means.
Ward's method
1. assigns cases to clusters randomly based oon basic criterion
2. measures the amount fo within-cluster variance (mean of squared difference of each case to cluster mean)
3. shifts a case to another initial cluster, if sum of squared mean is reduced, the case remains in cluster
4. repeats these steps until large number of combinations have been tested
These are the steps to conduct _____________.
k-means cluster analysis
If one or more of the variables are measured on a __________ scale, heirarchical clustering will likely be better. K-means clustering assumes all variables have been measured on at least a _________ scale.
nominal, ordinal
In ____________ clustering, cases can move from cluster to cluster as the analysis attempts to find the solution with the least amount of within cluster variance.
k-means
in _______ clustering it is not possible for a case to move to another cluster once assigned
heirarchical
K-means is based on ____________ as an indicator of homogeneity in which less _____________ = more homogeneity (both blanks are the same term)
within cluster variance
For cluster analysis it is recommended to have at least _________ cases. But this is under researched so may be inaccurate.
200
Sets found in the same folder