Quiz
Statistics and Research Design
This quiz has question types that open in the usual player.
Questions in this quiz (80)
1To test the hypothesis that violent movies make kids aggressive, [blank_start]violence[blank_end] would be the independent variable and [blank_start]aggression[blank_end] would be the dependent variable.
- Aviolence
- Baggression
2A mediator variable [blank_start]explains[blank_end] the relationship between IV and DVs. A moderator variable [blank_start]affects the strength of[blank_end] that relationship.
- Aexplains
- Baffects the strength of
3What would you use to measure a behaviour with no clear beginning or end (eg talking)?
- AInterval recording
- BEvent sampling
- CBehavior sampling
- DSituation sampling
4[blank_start]Content[blank_end] analysis organizes data into categories. [blank_start]Protocol[blank_end] analysis is a type of content analysis used in verbal problem solving studies. Behavioural sampling systematically records or rates quantitative aspects of behaviours and includes the following: [blank_start]Interval[blank_end] recording, for complex behaviours with no clear start/end points, and [blank_start]Event[blank_end] sampling, for counting less frequent behaviours. [blank_start]Situational[blank_end] sampling is used to observe across settings. [blank_start]Sequential[blank_end] analysis is used to identify more complex sequences of behaviour.
- AContent
- BProtocol
- CInterval
- DEvent
- ESituational
- FSequential
5A psychologist randomly selects 10 inpatient programs and then includes a random sample of patients from each program in her study on schizophrenia. This technique is called:
- AProbability sampling
- BSimple random sampling
- CStratified random sampling
- DCluster sampling
6An extraneous or confounding variable is a source of systematic error that should be controlled because it correlates with the ______________.
- ADependent variable
- BIndependent variable
- CSample
7Confounding variables are controlled in many ways, but NOT by:
- ARandom assignment
- BSelection of subjects with constant levels of these variables
- CPairing subjects on the basis of these variables
- D"Blocking" by including it as an additional DV
- EStatistical control via ANCOVA
8The standard error of the mean is calculated by dividing the [blank_start]population SD[blank_end] by the [blank_start]square root of the sample size[blank_end].
- Apopulation SD
- Bsquare root of the sample size
9The three predictions of the [blank_start]Central Limit Theorem[blank_end] are: 1) normality 2) equal means of sampling and population distributions and 3) SEM.
- ACentral Limit Theorem
10Under the normal curve, [blank_start]68[blank_end] % of scores fall within 1 SD of the M. [blank_start]95[blank_end] % fall within 2 SDs and [blank_start]99[blank_end] % fall within 3 SDs.
- A68
- B95
- C99
11Positive skew means the [blank_start]tail[blank_end] is positive and that the [blank_start]mean[blank_end] > [blank_start]median[blank_end] > [blank_start]mode[blank_end].
- Atail
- Bmode
- Cmean
- Dmedian
12A [blank_start]leptokurtic[blank_end] distribution is pointy, while a [blank_start]platykurtic[blank_end] distribution is flat.
- Aleptokurtic
- Bplatykurtic
13The level of significance is defined by the [blank_start]rejection[blank_end] region, also known as [blank_start]alpha[blank_end].
- Arejection
- Bretainment
- Calpha
- Dbeta
14A two tailed test must have a larger alpha region than a one tailed test.
- ATrue
- BFalse
15Drag and drop to complete the diagram
- A1 - alpha
- B1 - beta
- Calpha
- Dbeta
- Epower
- Ftype 2 error
- Gtype 1 error
16To increase statistical power:
- AIncrease alpha
- BDecrease sample size
- CIncrease effects of IV
- DMinimize error
- EUse a two tailed test
- FUse a parametric test
17Researchers consider [blank_start]power[blank_end] prior to decisions about the null hypothesis. In contrast, [blank_start]confidence[blank_end] refers to the certainty about a past decision and is higher when [blank_start]alpha is small[blank_end].
- Apower
- Bconfidence
- Calpha is small
- Dalpha is large
- Ebeta is small
- Fbeta is large
18ALL parametric (but NOT nonparametric) statistics require the following assumptions to minimize errors:
- ARandom selection
- BIndependence of observations
- CInterval or ratio scale
- DNormality in population distribution
- EHomoscedasticity (equal variances of populations represented by different groups)
19Parametric tests are most susceptible to violations of the following assumption:
- AHomoscedasticity
- BNormal distribution
- CInterval or ratio scale
- DIndependence of observations
20To evaluate a hypothesis about the shape of a distribution, you would use a [blank_start]nonparametric[blank_end] test. This type of test is [blank_start]less[blank_end] powerful than other tests and requires [blank_start]no[blank_end] assumptions about the shape of the population distribution.
- Anonparametric
- Bless
- Cno
21To choose a test, you must consider: 1) [blank_start]the scale of the DV[blank_end], 2) [blank_start]the number of IVs[blank_end], 3) [blank_start]if groups are independent or correlated[blank_end], 4) [blank_start]any confounds to control for[blank_end], and 5) [blank_start]the number of DVs[blank_end].
- Athe scale of the DV
- Bthe number of IVs
- Cif groups are independent or correlated
- Dany confounds to control for
- Ethe number of DVs
22Degrees of freedom determine the [blank_start]shape[blank_end] of the sampling distribution. To calculate df for a single sample chi squared test (or "goodness of fit test"), the formula is [blank_start]C-1[blank_end], which reflects the number of categories or levels. For a multiple sample chi squared test (or "contingency table"; with 2+ variables), df = [blank_start](C-1)(R-1)[blank_end]. For a t test, df = [blank_start]N-1[blank_end].
- Ashape
- BC-1
- C(C-1)(R-1)
- DN-1
23To count the number of relevant variables for a chi squared test, you do not have to distinguish between IVs and DVs.
- ATrue
- BFalse
24Chi square tests require:
- AFrequency data
- BNominal scales
- CIndependent observations
- DNormally distributed data
- EExpected frequencies of no less than five in any one category
- FExpected frequencies of no less than 10 in any one category
25Chi squared tests look at [blank_start]frequency[blank_end] data rather than the [blank_start]descriptive[blank_end] statistics used in other tests.
- Afrequency
- Bdescriptive
26Ordinal, or "rank ordered" data requires nonparametric alternative tests. A researcher would use the [blank_start]Mann-Whitney U Test[blank_end] instead of a t test with independent samples; the [blank_start]Wilcoxon Matched Pairs Signed Ranks Test[blank_end] instead of a t test for correlated samples; and the [blank_start]Kruskal-Wallis Test[blank_end] instead of a one way ANOVA. So, the [blank_start]Mann-Whitney[blank_end] would make sense for skewed but independent groups of IQ scores (DV). The [blank_start]Kruskal-Wallis[blank_end] would make sense for skewed (and therefore rank-converted) IQ scores across more than 2 levels of cognitive treatment. And the [blank_start]Wilcoxon[blank_end] would be applied to rank ordered data such as the IQ scores of matched pairs of patients across groups.
- AMann-Whitney U Test
- BWilcoxon Matched Pairs Signed Ranks Test
- CKruskal-Wallis Test
- DMann-Whitney
- EKruskal-Wallis
- FWilcoxon
27The [blank_start]experimentwise error[blank_end] is a factor in type 1 errors. Using a(n) [blank_start]ANOVA[blank_end] test instead of [blank_start]multiple t tests[blank_end] can help to reduce the number of comparisons in a single study so that this is minimized.
- Aexperimentwise error
- BANOVA
- Cmultiple t tests
28To compare the mean of a single group to the known mean of a population (acting here as a no-treatment control group), you would use a t test [blank_start]for a single sample[blank_end] and df = [blank_start]n-1[blank_end]. To compare the means of two independent groups, you would use a t test [blank_start]for independent samples[blank_end] and df = [blank_start]n-2[blank_end]. To compare the means of two correlated groups (eg same subjects over time), you would use a t test [blank_start]for correlated samples[blank_end] and df = [blank_start]number of matched pairs - 1[blank_end].
- Afor a single sample
- Bn-1
- Cfor independent samples
- Dn-2
- Efor correlated samples
- Fnumber of matched pairs - 1
29T tests assume specific characteristics of the sampling distribution of means, therefore they require interval or ratio scaled data.
- ATrue
- BFalse
30To compare two means in a within subject design (eg twins assigned to different treatment groups), which would be the appropriate test?
- AANOVA
- BT test for correlated samples
- CKruskal-Wallis test
- DChi squared test
31If the IV includes more than two levels (and all standard assumptions are true), this would indicate the use of a(n) [blank_start]ANOVA[blank_end] instead of a(n) [blank_start]t test[blank_end].
- AANOVA
- Bt test
32T tests are only used to compare two means.
- ATrue
- BFalse
33ANOVA is used to compare 2 or more means. Its advantage is that it minimizes experimentwise error by holding [blank_start]type 1 error[blank_end] constant at a set level.
- Atype 1 error
34The difference between a one way ANOVA versus a factorial ANOVA is in the number of [blank_start]IVs[blank_end]. A factorial ANOVA examines [blank_start]interaction[blank_end] effects, where a one way ANOVA only examines a [blank_start]main[blank_end] effect.
- AIVs
- Binteraction
- Cmain
35ANOVA partitions variability by examining [blank_start]sums of squares[blank_end]. In a one way ANOVA, [blank_start]SST = SSB + SSW[blank_end]. SSB in this equation refers to the variability [blank_start]between groups[blank_end], while SSW refers to the variability [blank_start]within them[blank_end].
- Asums of squares
- Bwithin them
- CSST = SSB + SSW
- Dbetween groups
36F = [blank_start]MSB/MSW[blank_end] = (treatment + [blank_start]error[blank_end])/error
- AMSB/MSW
- Berror
37MST = [blank_start]SST/df[blank_end], MSB = [blank_start]SSB/df[blank_end] and MSW = [blank_start]SSW/df[blank_end]. These equations are relevant to the following statistical test: [blank_start]ANOVA[blank_end].
- ASSW/df
- BANOVA
- CSST/df
- DSSB/df
38[blank_start]MSW[blank_end] is a measure of variability within groups of subjects who have received the same treatment, therefore it is assumed to be due to [blank_start]error[blank_end] only. [blank_start]MSB[blank_end] is a measure of variability between groups of subjects that have been treated differently, so it is thought to reflect both error and treatment effects.
- AMSW
- Berror
- CMSB
39If the null hypothesis is true, F = 0.
- ATrue
- BFalse
40As F increases, so does it's chance of being statistically significant.
- ATrue
- BFalse
41The test statistic for ANOVA is:
- AF
- Bt
- Cr
- DD
42A significant ANOVA requires post hoc tests when it involves 3+ groups.
- ATrue
- BFalse
43Which tests would you consider to follow up a significant one way ANOVA?
- ATukey
- BScheffe S
- Cpost hoc tests
- DWilcoxon
- EMann-Whitney
- FHomoscedasticity
44The statistics for the following tests are as follows: Mann-Whitney [blank_start]U[blank_end], Wilcoxon [blank_start]T[blank_end], and Kruskal-Wallis [blank_start]H[blank_end].
- AU
- BT
- CH
45Which of the following is a "factorial" ANOVA?
- AOne way ANOVA
- BThree way ANOVA
- CANOVA with 2 IVs
- DTwo way ANOVA
46Both types of these factorial ANOVAs increase statistical power by reducing within group variability due to an extraneous variable. The [blank_start]randomized block ANOVA[blank_end] is used in a preplanned way, by treating the extraneous variable as [blank_start]another IV[blank_end]. The [blank_start]ANCOVA[blank_end] may be used used after group assignment, by using [blank_start]multiple regression[blank_end] to [blank_start]adjust DV scores[blank_end] according to the extraneous variable.
- Arandomized block ANOVA
- Bmultiple regression
- Canother IV
- DANCOVA
- Eadjust DV scores
47You would use a [blank_start]repeated measures ANOVA[blank_end] for a within subjects study that involves more than 2 IV conditions and interval DV data. If the same study involved only 2 conditions, you would use a [blank_start]t test for correlated samples[blank_end].
- Arepeated measures ANOVA
- Bt test for correlated samples
48A study looks at the impact of dose (high, med and low) and time (immediate, 2 mos, 6 mos) on a symptom test score. Participants are assigned to dosage group and tested at each of the three time points. The appropriate test for this study is a [blank_start]mixed ANOVA[blank_end] because it makes comparisons [blank_start]both within and between[blank_end] groups. Another name for this test is a [blank_start]split-plot ANOVA[blank_end].
- Amixed ANOVA
- Bboth within and between
- Csplit-plot ANOVA
49F = [blank_start]t[blank_end] when only two levels are included in a one way ANOVA.
- At
50To evaluate whether or not there is a significant linear or nonlinear relationship between two variables, you could use [blank_start]trend analysis[blank_end] to look at the shape of this relationship.
- Atrend analysis
51To assess the effects of an IV on 2+ DVs simultaneously, you could use a [blank_start]MANOVA[blank_end].
- AMANOVA
52[blank_start]Effect size[blank_end] indicates the practical impact or "strength" of the results obtained.
- AEffect size
53Two measures of effect size are 1) [blank_start]Cohen's d[blank_end], which measures the difference between two groups (mean of group1 - mean of group2 / pooled sd of both groups) and 2) [blank_start]Eta squared[blank_end], which indicates the percent of variance in DV that is accounted for by the variance in IV.
- ACohen's d
- BEta squared
54For effect sizes measured by d, a small effect = [blank_start]0.2[blank_end], a medium effect = [blank_start]0.5[blank_end] and a large effect = [blank_start]0.8[blank_end].
- A0.2
- B0.5
- C0.8
55If eta squared (n^2) = 0.55, this means that ____ % of the variability in Y is explained by the variability in X.
- A55
- B30.25
- C7.5
- D45
56To correlate two interval or ratio variables, [blank_start]Pearson r[blank_end] is used. For two rank ordered variables, [blank_start]spearman rho[blank_end]. For two nominal variables, [blank_start]contingency[blank_end]. For two true dichotomies, [blank_start]phi[blank_end]. For two artificial dichotomies, [blank_start]tetrachoric[blank_end]. A [blank_start]point biserial[blank_end] correlation coefficient is applied for a true dichotomy plus an interval or ratio variable. If this dichotomy is [blank_start]artificial[blank_end], it's [blank_start]just biserial[blank_end]. Finally, when the relationship between variables is nonlinear but both are at least interval scale, we use [blank_start]eta[blank_end].
- APearson r
- Bspearman rho
- Ccontingency
- Dphi
- Etetrachoric
- Fpoint biserial
- Gartificial
- Hjust biserial
- Ieta
57Most correlation coefficients require the following assumptions:
- ALinearity
- BUnrestricted range
- CHomoscedasticity
- DRandom selection
- EIndependent observations
- FInterval scale
58For parametric tests, this refers to the assumption of equal variances in the relevant population groups. For correlation coefficients, this refers to the assumption of roughly equal variances for Y at different levels of X. When violated, the scatter plot looks like a triangle. What is [blank_start]homoscedasticity[blank_end]?
- Ahomoscedasticity
59The range of Y scores at every value of X is equal to the total range of Y scores. Therefore, r = [blank_start]0[blank_end].
- A0
60The "[blank_start]coefficient of determination[blank_end]" is another name for the shared [blank_start]variability[blank_end] between two variables that is indicated by r squared. The exception to the rule for squaring r is when it is a [blank_start]reliability[blank_end] coefficient, which reflects a correlation of a measure with itself. In this case, you interpret it directly as a measure of "[blank_start]true score[blank_end] variability" (rxx instead of rxy).
- Acoefficient of determination
- Bvariability
- Creliability
- Dtrue score
61When the alpha is .05 and n = 10, r must be at least .63 to reach significance. When n is 50, the requirement for a significant correlation would be:
- AHigher
- BLower
- CThe same
- DIndependent of n
62Regression analysis assumes a linear relationship between X and Y.
- ATrue
- BFalse
63To find the "line of best fit", we use the method of [blank_start]least squares[blank_end].
- Aleast squares
64The predictive accuracy of a regression equation is directly related to the magnitude of the correlation coefficient.
- ATrue
- BFalse
65[blank_start]Multivariate[blank_end] techniques typically investigate the [blank_start]linear[blank_end] relationships between [blank_start]3[blank_end] or more variables. Dependence methods are used when there is [blank_start]a distinction[blank_end] between IVs and DVs and interdependence methods are used when there is [blank_start]no distinction[blank_end].
- AMultivariate
- Blinear
- Ccurvilinear
- Dnonlinear
- E3
- F2
- Ga distinction
- Ha relation
- Ino distinction
- Jno relation
66To choose a multivariate technique, you should consider the scale of [blank_start]the predictor and criterion[blank_end] variables and the number of [blank_start]criterion[blank_end] variables.
- Athe predictor and criterion
- Bcriterion
67Multiple regression requires:
- AA continuous criterion
- BA continuous predictor
- CTwo or more variables
- DMulticollinearity
68The output of a multiple regression analysis is a multiple [blank_start]correlation coefficient[blank_end] and a multiple [blank_start]regression equation[blank_end]. R indicates the degree of association between the criterion and a [blank_start]linear combination of[blank_end] predictors. To measure shared variability, you would [blank_start]square[blank_end] R.
- Acorrelation coefficient
- Bregression equation
- Clinear combination of
- Dsquare
69Y' = b1X1 + b2X2 + ... + bkXk + a. In this equation, Y' is the [blank_start]predicted score on Y[blank_end]. X refers to [blank_start]obtained scores on X[blank_end]. b refers to [blank_start]regression coefficients[blank_end]. a is [blank_start]the Y intercept[blank_end].
- Apredicted score on Y
- Bactual score on Y
- Cobtained scores on X
- Ddiscrete variables
- Enumber of measurements
- Fregression coefficients
- Gcorrelation coefficients
- Hlines of best fit
- Ithe Y intercept
- Jthe X intercept
- Kthe critical value
- Lthe criterion variable
70In the multiple regression equation, the magnitude of b is determined by the correlation between the predictor and the criterion.
- ATrue
- BFalse
71A high correlation between two predictors in a multiple regression equation is referred to as [blank_start]multicollinearity[blank_end]. It is a [blank_start]problem[blank_end] because it means that the predictors are [blank_start]redundant[blank_end] and the magnitude of b is [blank_start]not proportional to[blank_end] the true correlation.
- Amulticollinearity
- Bproblem
- Credundant
- Dnot proportional to
72The simplest form of multiple regression is [blank_start]simultaneous[blank_end] regression. In contrast, [blank_start]stepwise[blank_end] regression [blank_start]adds or subtracts[blank_end] predictors in subsequent steps in order to explain the most variability while using [blank_start]the fewest[blank_end] predictors.
- Asimultaneous
- Bstepwise
- Cadds or subtracts
- Dthe fewest
73A researcher might opt for multiple regression instead of ANOVA because:
- AGroups are unequal in size
- BIVs are continuously measured and ANOVA would convert them to categories
- CTo improve power
- DTo improve robustness
- ETo determine the best subset of predictors
- FGroups are equal in size
- GIVs are ratio scale
- HMultiple DVs
74Applying a multiple regression equation to a new sample is called [blank_start]cross validation[blank_end]. It improves [blank_start]external[blank_end] validity, but is associated with [blank_start]shrinkage[blank_end] (where the size of the correlation coefficient decreases) - especially when the original sample was [blank_start]small[blank_end] and the number of predictors is [blank_start]large[blank_end].
- Across validation
- Bexternal
- Cshrinkage
- Dsmall
- Elarge
75Basic multiple regression uses [blank_start]2+ continuous or discrete predictors[blank_end] for [blank_start]a single continuous criterion[blank_end]. If all predictors are continuous, you could use [blank_start]canonical correlation[blank_end] to predict status on [blank_start]2+ continuous predictors[blank_end] or [blank_start]discriminant function analysis[blank_end] to predict status on [blank_start]a single nominal criterion[blank_end]. The accuracy of this latter technique is assessed by determining the "[blank_start]hit rate[blank_end]". To predict status on [blank_start]a single discrete criterion[blank_end] without assuming linearity, you would opt for [blank_start]logistic regression[blank_end], which requires 2+ continuous or discrete predictors.
- A2+ continuous or discrete predictors
- Ba single continuous criterion
- Ccanonical correlation
- D2+ continuous predictors
- Ediscriminant function analysis
- Fa single nominal criterion
- Ghit rate
- Ha single discrete criterion
- Ilogistic regression
76Causal modelling tests predefined models or theories and suggest but cannot prove causality.
- ATrue
- BFalse
77To translate a causal model about the strength and direction of one way relationships between variable pairs, you could use [blank_start]path analysis[blank_end]. If the relationships are two-way (or [blank_start]recursive[blank_end]), you would choose [blank_start]LISREL[blank_end], which stands for [blank_start]linear structural relations analysis[blank_end]. LISREL is also good because it considers the hypothesized [blank_start]latent traits[blank_end] of variables as well as [blank_start]measurement[blank_end] error.
- Apath analysis
- Bmeasurement
- Cvariability
- Dmaturation
- Ecross validation
- Fdiscriminant function analysis
- Gcodependent
- Hcluster analysis
- Icanonical correlation
- Jlinear systems and relations analysis
- Krecursive
- LLISREL
- Mlinear structural relations analysis
- Nlatent traits
78Multivariate analysis is mainly used for prediction, causal [blank_start]modelling[blank_end] and data [blank_start]reduction[blank_end].
- Amodelling
- Bexplanation
- Creduction
- Danalysis
79Classifying a large group of variables into smaller subgroups is called [blank_start]data reduction[blank_end].
- Adata reduction
80[blank_start]Factor analysis[blank_end] is used to develop subscales for tests. [blank_start]Cluster analysis[blank_end] is used to define data driven subgroups of similar people or objects. In contrast, [blank_start]discriminant function analysis[blank_end] and logistic regression sort people into predefined groups.
- AFactor analysis
- BCluster analysis
- Cdiscriminant function analysis
Information
- Description
- Quiz on Statistics and Research Design, created by N O on 07/05/2017.
- Author
- N O
- Language
- English
- Views
- 21
- Tags
- Rating
- No ratings yet Rate this quiz