Examtests with Multivariate Data Analysis Text Book by Leiden University

Which analysis method can be used for separate types of problems? - ExamTests 0

Questions

Question 1

A researcher aims to predict whether FC Barcelona will win the competition this season based on the variables 'invested money' and 'talent of the selection'. What method do we use?

1. Simple regression analysis.

2. Multiple regression analysis.

3. Logistic regression analysis.

4. ANOVA.

Question 2

One person is better than the other in postponing reward. On age three, we can already see this in performance on the marshmallow test. A developmental psychologist believes that children who move a lot in the uterus, who cry loud at birth, and who are very passionate, will often fail the marshmallow test. She measures the movement at 6 months in the uterus (10-point scale), the volume of crying at birth (in dB) and the number of tantrums at the age of 14 months in 200 participants. Then, each child does the marshmallow test at age 3, in which only not eating the marshmallow is registered.

What is the best technique to use when analyzing the data?

1. MRA.

2. ANOVA.

3. ANCOVA.

4. LRA.

Question 3

During summer, the police often receive more complaints than in other seasons. In order to arrive at a targeted prevention policy, the noise hindrance per house is measured by the number of decibels produced at 300 homes in the same district. An attempt is made to predict this noise hindrance as accurately as possible on the basis of sex, age (in years), number of children, extraversion (on a 20-point scale) and IQ of the main resident. What is the most suitable technique for this?

1. MRA.

2. ANOVA.

3. ANCOVA.

4. LRA.

Question 4

Research has shown that a disproportionate number of Peruvian footballers have a very good ball technique. A possible explanation is that Peruvian street footballers usually play with inferior, mainly ovoid, balls and can only use them with a fabulous technique.
For the time being, this statement is based solely on informal observations. A researcher decides to check whether the balls are less round in Peruvian street football than in street football in other countries. To this end, he chooses five countries (Colombia, Peru, Switzerland, Belgium and Iran) and sends an experienced ballometrician (who knows nothing about the purpose of the research) to each of these countries. In each country, at 100 randomly selected street soccer games, the measures of the ball during the break, leading to a score from 0 (perfectly round) to 100 (absolutely not round) per ball. The hypothesis is that the Peruvian balls are less round than the balls from the other countries.

What is the most suitable technique to test this hypothesis?

1. MRA.

2. ANOVA.

3. ANCOVA.

4. LRA.

Question 5

Much pseudo-research has shown that men provide better education than women, but many people disagree. They think men are better judged because men are taller on average and have heavier votes. To verify this, New York Psychology teachers measure the average student opinion about their educational qualities, the heaviness of the voice (on a 20-point scale), the height (in millimeters) and sex (male or female). The prediction is that the male teachers will on average be rated better than the female teachers, but this difference will disappear after correction for body length and heaviness of the voice.

What is the most appropriate technique to investigate this prediction?

1. MRA.

2. ANOVA.

3. ANCOVA.

4. LRA.

Question 6

An insurance psych

A researcher aims to predict whether FC Barcelona will win the competition this season based on the variables 'invested money' and 'talent of the selection'. What method do we use?

1. Simple regression analysis.

2. Multiple regression analysis.

3. Logistic regression analysis.

4. ANOVA.

ologist, specialized in behavioral aspects of DIY accidents, finds that in many of the DIY incidents investigated, the seriously injured handyman had not looked into the tool manual. It had been looked at fleetingly, but only rarely studied carefully. The psychologist wonders what factors determine how long the handyman (M / F) studies the manual: type of tool, sex of the handyman, number of years of work experience.

In collaboration with the PRAXIS, the psychologist has 400 buyers of a glue gun, multi-tool or sander complete a survey one month after purchase. This contains questions about the type of tool (glue gun, multi-tool, or sander), sex, number of years of DIY experience, and number of minutes that the manual for the purchased device was read.

What is the most suitable technique to answer the above question?

1. MRA.

2. ANOVA.

3. ANCOVA.

4. Mediation analysis.

Question 7

A positive psychologist suspects that introverts enjoy doing odd jobs less than introverts, because they are more in a “flow” state during odd jobs. To investigate this she brings 300 experienced handymen to the lab, in which they all do the same job with 38 electrodes on their heads. Beforehand, everyone measures the degree of introversion, during the job the number of seconds in flow is measured based on flow-related brain patterns, and afterwards everyone is asked on a 9-point scale how much fun they had in the job.

What is the most suitable technique to investigate the above presumption?

1. ANOVA.

2. ANCOVA.

3. Repeated measures ANOVA.

4. Mediation analysis.

Question 8

A director of a roofing company asks to what extent the following matters determine how long a job lasts: the size of the roof, the number of used roofing materials, the number of years of experience of the roofing materials? To answer this question, it obtains a number of data from the business administration of all the jobs over the past five years: roof area (in square meters), number of roofing tiles used, average number of years of experience of the roofing tiles used and number of hours between the beginning and end of the job.

What is the most suitable technique for answering this research question?

1. MRA.

2. ANOVA.

3. ANCOVA.

4. MANOVA.

Question 9

When a basketball game ends equally, it must sometimes be decided by taking penalty throws. According to a well-known Dutch coach, it makes little sense to train specifically for this, but some think differently. A research group suspects the following: the greater the overall productivity of a team and the more training on penalty throws, the greater the chance of winning a penalty throw series.
It is decided to investigate this by measuring the following of all Dutch games from 1973 to the present that have been decided on penalties for each home team.

• Penalty throwing training: how many minutes had they been trained on penalties in the previous week?
• General productivity: how many goals per game had the team scored on average in the previous five games?
• Result: was the penalty shootout won or lost?

What is the most suitable technique to investigate the above presumption?

1. MRA.

2. ANOVA.

3. ANCOVA.

4. LRA.

Question 10

A political analyst wonders which president has the most charisma in the eyes of the American voters. To this end, he asks a large, representative sample of voters to rate the charisma of five (former) presidents using an 11-point scale ranging from 0 = “no charisma at all” to 10 = “great charisma”.

What is the most appropriate technique for comparing the charism of the five presidents?

1. MRA.

2. ANCOVA.

3. MANOVA.

4. Repeated measures ANOVA (RMA).

Answer indication

1. C. Logistic regression analysis.

2. D. LRA.

3. A. MRA.

4. B. ANOVA.

5. C. ANCOVA.

6. C. ANCOVA.

7. D. Mediation analysis.

8. A. MRA.

9. D. LRA.

10. D. Repeated measures ANOVA (RMA).

How does multiple regression analysis (MRA) work? - ExamTests 1

Questions

Question 1

When do we speak of multicollinearity in multiple regression analysis (MRA)?

1. When two or more predictors are highly correlated.

2. When the measurement errors of several predictors are highly correlated.

3. When the residuals of the regression model are correlated.

4. When one predictor is highly correlated with the dependent variable.

Question 2

With what statistic do we measure whether adding specific a predictor variable significantly improves our model?

1. R².

2. R²_change.

3. Std. error.

4. Adjusted R².

Question 3

Why can't we use Pearson correlation (R) in the case of a categorical variable with more than two categories?

1. Because this variable is not a ratio or interval variable.

2. Because the variables cannot have negative or positive values.

3. Because classification is not arbitrary.

Question 4

Given is the following SPSS table. What procedure was used?

1. The standard method.

2. The hierarchical procedure.

3. The 'removal' procedure.

Question 5

What variable in the same table explained the dependent variable best?

1. Variable 1.

2. Variable 2.

3. Variable 3.

4. Variable 4.

Question 6

Is there a variable that could better be not added to the model?

1. Yes, the variable that is added at step 4 does not significantly improve the model.

2. No, but only because the significance level is still below .05.

3. No, because the explained variance and F-values have increased at each step.

4. We cannot be sure based on the information in the table.

Question 7

What criteria are important when evaluating a model?

1. The amount of variance explained and the degree of multicollinearity.

2. The degree of multicollinearity and the statistical significance.

3. The statistical significance and the amount of explained variance.

4. The amount of explained variance, statistical significance, and the degree of multicollinearity.

Question 8

In a regression analysis with data of 150 participants the regression weight of X₂ is not significant, but the correlation is however high: .6. The correlation between X₁ and Y is .8, and the correlation between X₁ and X₂ is .8. What is the explanation for this?

1. There is an interaction effect between X₁ and X₂.

2. The data is non-linear.

3. X₁ does not have a unique contribution.

4. X₂ does not have a unique contribution.

Question 9

Given is the following SPSS table. What is the standardized equation?

1. Y= -18.534 + 7.890*X1 – 3.980*X2.

2. Y= 7.890*X1 – 3.980*X2.

3. Y= -18.534 + 0.673*X1 – 0.380*X2.

4. Y= 0.673*X1 – 0.380*X2.

Question 10

What is the meaning of B in the table of question 9?

1. When X₁ increases with 1 unit, Y will increase with 7.890.

2. When X₁ increases with 7.890, Y will increase with 1 unit.

3. When X₁ increases with 1 unit, Y will increase with 7.890*.673.

Question 11

What is not a factor that can strongly influence Pearson r and the regression coefficients?

1. Limited variance.

2. Standard measurement errors.

3. Outliers.

4. Non-linearity.

Question 12

Multicollinearity leads to a higher/lower R² and is not good/good when we want to understand the interaction between variables.

1. Higher/good.

2. Higher/not good.

3. Lower/good.

4. Lower/not good.

Question 13

With which one of these methods are predictors added stepwise?

1. Forward selection procedure.

2. Backward selection procedure.

3. Standard selection procedure.

Question 14

In a model there are two partial correlations with the dependent variable: r_x1y = .6 and r_x2y = .4. The explained variance is 60%. How much variance is explained by both of these variables?

1. 4%.

2. 6%.

3. 8%.

4. We cannot be sure based on this information.

Question 15

A health psychologists investigates with an MRA whether OCD can be predicted by the amount of rumination and the amount of doubt. This yields the following output:

What is the total proportion of variance explained of OCD?

1. .120.

2. .528.

3. .647.

4. .766.

Question 16

A professor researches with a MRA if the grade for MVDA can be predicted by the number of study hours per week, the number of travelling hours, and the grade for Psychometrics. This yields the following output:

What is according to this regression model the predicted grade for MVDA of a student who studies 5 hours a week, travels 4 hours a week, and got an 8 for Psychometrics?

1. 3.4.

2. 4.0.

3. 6.6.

4. 7.1.

Answer indication

1. A. When two or more predictors are highly correlated.

2. B. R²_change.

3. C. Because classification is not arbitrary.

4. B. The hierarchical procedure.

5. A. Variable 1.

6. C. No, because the explained variance and F-values have increased at each step.

7. C. The statistical significance and the amount of explained variance.

8. D. X₂ does not have a unique contribution.

9. D. Y= 0.673*X1 – 0.380*X2.

10. A. When X₁ increases with 1 unit, Y will increase with 7.890.

11. B. Standard measurement errors.

12. B. Higher/not good.

13. A. Forward selection procedure.

14. C. 8%.

15. C. .647.

16. C. 6.6.

How does Analysis of Variance (ANOVA) work? - ExamTests 2

Questions

Question 1

When do we speak of homogeneity in analysis of variance (ANOVA)?

1. When all independent variables have the same variance.

2. When the dependent variable is normally distributed.

3. When the variance of the dependent variable is the same in all groups.

4. When all groups have the same size.

Question 2

When do we speak of a significant interaction effect?

1. When the independent variables are significantly correlated with each other.

2. When the effect of the independent variable can be explained by another variable that was not included in the model.

3. When the effect of one independent variable significantly differs for different categories of another independent variable.

4. When two independent variables significantly predict the dependent variable.

Question 3

From an ANOVA table we already have the following values: SS_{between groups} = 400, SS_{within groups} = 150, df_{between groups} = 4, df_{within groups} = 16. What is the F-value?

1. 2.667.

2. 4.

3. 10.667.

4. 25.

Question 4

In a ... sample, a (n) ... correlation easily becomes significant.

1. Big, average.

2. Big, low.

3. Small, high.

4. Small, low.

Question 5

Which statement is true?

I. The strength of a certain effect in ANOVA is indicated by R².
II. The F-value in an ANOVA table is calculated by dividing the mean SS_model by the mean SS_error.

1. I is true.

2. II is true.

3. Both are true.

4. Neither are true.

Question 6

A consumer psychologist performs a between-subjects ANOVA to study the effect of personality type (introvert, extrovert, ambivert) and gender (female, male) on consumption. When do we speak of an interaction between personality type and gender?

1. If all six groups have the same size.

2. If within each personality type men buy more than women.

3. If the effect of gender is the same for each personality type.

4. If the effect of personality type is different for men than for women.

Question 7

What kind of information does a significant Levene's test give us?

1. That an independent variable is a significant predictor of a dependent variable.

2. That the variances aren't equally distributed.

3. That one independent variable influences the categories of another independent variable.

Question 8

What table shows us information about the underlying effects between variables after running an ANOVA?

1. The coefficients table.

2. The correlations table.

3. The Pairwise Comparisons table.

Question 9

When do we speak of homoscedasticity in ANOVA?

1. When the dependent variable has the same variance in all groups.

2. When the independent variables all have the same variance.

3. When the dependent variable is normally distributed.

4. When the groups contain the same number of participants.

Question 10

A researcher investigates what the effect of a study book (English for beginners / I learn English! / Magic English) and gender (boy / girl) is on performance. This yields the following output:

The psychologists investigates the hypothesis with a between-subjects ANOVA design. This yields the two following statements about the robustness of the F-tests:

I. The F-tests are robust against violation of normality.
II. The F-tests are robust against violation of homogeneous group variances.

Which statement is true?

1. I is true.

2. II is true.

3. Both are true.

4. Neither are true.

Question 11

In a balanced study with 88 participants a social psychologist investigates the effect of gender (female / male) and age (4 categories) on tablet use with a between-subjects ANOVA. This yields the following table, which is partly filled in:

Complete the table. What effect has the greatest F-value?

1. Gender.

2. Age.

3. Gender*Age.

4. All F-values are equally as great.

Answer indication

1. C. When the variance of the dependent variable is the same in all groups.

2. C. When the effect of one independent variable significantly differs for different categories of another independent variable.

3. C. 10.667.

4. B. Big, low.

5. B. II is true.

6. D. If the effect of personality type is different for men than for women.

7. B. That the variances aren't equally distributed.

8. C. The Pairwise Comparisons table.

9. A. When the dependent variable has the same variance in all groups.

10. B. II is true.

11. A. Gender.

How does Analysis of Covariance (ANCOVA) work? - ExamTests 3

Questions

Question 1

Suppose we perform an analysis of covariance (ANCOVA) with one factor and one covariate. It turns out that the groups do not differ on the covariate. Furthermore, the within-groups regression coefficient bw is significantly smaller than 0. What goal would performing an ANCOVA have in this case, compared to an ANOVA without the covariate?

1. No reduction of error variance and no removal of systematic bias.

2. Removal of systematic bias only.

3. Reduction of error variance only.

4. Both reduction of error variance and removal of systematic bias.

Question 2

Why should one use covariates sparingly in ANCOVA?

1. The statistical power of the F test decreases if the covariate hardly correlates with the dependent variable.

2. The F test becomes less robust to violations of homoscedasticity if too many covariates are used.

3. The F value increases if the covariate hardly correlates with the dependent variable.

4. The explained variance may drop if too many covariates are used.

Question 3

Which statement is true?

I. Covariance indicates how the differences on one variable relate to differences on another variable.
II. In case of a perfect positive relationship, the points of a scatterplot display a straight line downwards.

1. I is true.

2. II is true.

3. Both are true.

4. Neither are true.

Question 4

What is in ANCOVA a reason why not to use to use to many covariates in a model?

1. The statistical power of the F-test will go down if an added covariate does not or barely correlates with the dependent variable.

2. The F-value will increase when an added covariate does not or barely correlate with the dependent variable.

3. In the case of too many covariates, the F-test will be less robust against violation of homoscedasticity.

4. With too many covariates, the explained variance will decrease.

Question 5

In an ANCOVA with one factor and one covariate the groups appear not to be differing on the covariate. Also, the within-groups regression weight bw is significantly smaller than 0. What is the goal of ANCOVA in this case, compared to ANOVA without the covariate?

1. Reduction of error variance and elimination of systematic bias.

2. Only reduction of error variance.

3. Only elimination of systematic bias.

4. Neither of these.

Question 6

A health psychologists investigates the effect of three different interventions targeted at quitting smoking. The three interventions are called "Don't do it!", "Stop coughing", and "Be free". For each of the interventions, 50 participants are recruited. The psychologist measures the number of smoked cigarettes a day after the intervention.

There seems to be a difference between the interventions in the mean number of kCAL that the participants consume each day. Consuming calories negatively correlates with smoking. The psychologist decides to run an ANCOVA with the intervention as a factor, consumption as covariate, and smoking as the dependent variable. The pooled regression weight between smoking and consumption (bw) is -.10.

The three interventions have the following means on the covariate and the dependent variable:

After which intervention do the participants smoke the least number of cigarettes after correction for consumption?

1. Don't do it!

2. Stop coughing.

3. Be free.

4. After correction for consumption all interventions have the same mean.

Answer indication

1. C. Reduction of error variance only.

2. C. The F value increases if the covariate hardly correlates with the dependent variable.

3. A. I is true.

4. A. The statistical power of the F-test will go down if an added covariate does not or barely correlates with the dependent variable.

5. A. Reduction of error variance and elimination of systematic bias.

6. B. Stop coughing.

How does MANOVA work? - ExamTests 4

Questions

Question 1

What is Box's M used for in MANOVA?

1. To test the assumption of homogeneity of the variance-covariance matrices.

2. To test whether the groups differ significantly from each other.

3. To test the assumption of independent errors.

4. To test the significance of the discriminant functions.

Question 2

When is it not a good idea to run an MANOVA?

1. When the independent variables are highly correlated.

2. When you want to investigate multiple categories of the independent variable.

3. When the dependent variables are averagely correlated.

4. When you want to identify which dependent variable causes the most variance.

Question 3

In a study, three dependent variables are studied in two groups. Group 1 contains 20 persons and group 2 contains 30 persons. The Hotelling's T2 value is 3.7249. What is the F-value?

1. 1.024.

2. 1.189.

3. 11.024.

4. 11.661.

Question 4

What is the maximum number of discriminant function variates for a MANOVA with three groups and four dependent variables?

1. 1.

2. 2.

3. 3.

4. 4.

Question 5

We have the following given statistics of a model: rxiy = .34, multiple correlation R = .56, and VAF = .31. What is the structure coefficient?

1. .912.

2. 1.097.

3. 1.647.

4. .607.

Question 6

What statement is true about evaluating individual predictors?

1. Evaluating individual predictors is useful when the model does not predict better than chance level.

2. Both b-weights and beta-weights can be used to evaluate individual predictors.

3. Beta-weights can be influenced by the variability of a variable, extra predictors, and measurement errors.

4. Evaluating beta-weights and structure coefficients can be sufficient.

Question 7

Which statement is true?

I. An advantage of a multivariate factor design is that it shows how the independent variables interact in order to have an influence on the dependent variable.
II. The value of Wilk's lambda shows us the proportion of unexplained variance.

1. I is true.

2. II is true.

3. Both I and II are true.

4. Neither I or II is true.

Question 8

What is a vector?

1. A variable that has an influence on multiple dimensions.

2. A variable that can't be manipulated.

3. A weighted sum of dependent variables that together explain a phenomena.

Question 9

A health psychologist investigates whether three groups can be distinguished on the basis of four types of physical traits with a discriminant analysis. This yields the following output:

Structure matrix

Functions at Group Centroids

Which statement is true?

I. The first discriminant function variate is mainly determined by x₃ and x₄. This function primarily distinguishes group A from C.
II. The second discriminant function variate is mainly determined by x₂ and x₄. This function primarily distinguished group B from C.

1. I is true.

2. II is true.

3. Both I and II are true.

4. Neither I or II is true.

Answer indication

1. A. To test the assumption of homogeneity of the variance-covariance matrices.

2. A. When the independent variables are highly correlated.

3. B. 11.024.

4. B. 2.

5. D. .607.

6. C. Beta-weights can be influenced by the variability of a variable, extra predictors, and measurement errors.

7. C. Both I and II are true.

8. C. A weighted sum of dependent variables that together explain a phenomena.

9. C. Both I and II are true.

How does repeated ANOVA measurements work? - ExamTests 5

Questions

Question 1

Evaluate the following two statements on the role of contrasts in the multivariate approach to repeated measures ANOVA with p variables.

I. The multivariate tests are based on a set of (p-1) linear independent contrasts
II. The contrasts allow testing specific hypotheses about mean differences between groups.

1. Both I and II are true.

2. I is true.

3. II is true.

4. Neither I or II is true.

Question 2

With what participants design is there a treatment-effect?

1. A 2x2 design.

2. A one-way between-subjects design.

3. A repeated measures design.

4. A two-way simple mixed design.

Question 3

Judge the following statements about the role of contrasts in the multivariate approach of a repeated measures ANOVA with p variables.

I. The multivariate tests are based on a set of p-1 independent contrasts.
II. The contrasts make it possible to test specific hypotheses about the differences in group means.

1. Both I and II are true.

2. I is true.

3. II is true.

4. Neither I or II is true.

Question 4

Given are the following four contrasts for five variables:

Which of the following pairs of contrasts is not orthogonal?

1. L₁, L_2.

2. L₂, L_3.

3. L₁, L_4.

4. L₃, L_4.

Question 5

What is the relative importance of the three contrast for this group, from most important to the least important?

1. L₁ - L₂ - L_3.

2. L₁ - L₃ - L_2.

3. L₃ - L₂ - L_1.

4. L₃ - L₁ - L_2.

Answer indication

1. A. Both I and II are true.

2. C. A repeated measures design.

3. A. I and II are both true.

4. D. L₃, L_4.

5. D. L₃ - L₁ - L_2.

How does Logistic Regression Analysis (LRA) work? - ExamTests 6

Questions

Question 1

An educational psychologist performs a logistic regression analysis (LRA) to investigate if passing an exam can be predicted from the number of plenary lectures that were attended. The odds ratio of the predictor (number of attended lectures) equals 5. What does this mean?

1. For every exam score the probability of passing is five times as large as the probability of failing.

2. For every exam score the odds of passing are five times as large as the odds of failing.

3. The probability of passing becomes five times as large if you attend one more lecture.

4. The odds of passing become five times as large if you attend one more lecture.

Question 2

X1 has a regression coefficient of .568. The model has a constant of -3.734. How great is the probability that an individual belongs to the target group?

1. .08.

2. .25.

3. .40.

4. .04.

Question 3

Given is the classification table of a research towards whether a treatment was successful or not. What percentages should be filled in at A and B?

1. A: 66, B: 45.

2. A: 45, B: 55.

3. A: 30, B: 45.

4. A: 66, B: 64.3.

Question 4

A researcher wants to run a LRA on his data from a sample of 100 participants. The predictor variables are on an interval level, and the dependent variable is dichotomous. The researchers find that some of his predictor variables are strongly correlated, the predictors are linearly related to the log odds, and the errors are independent. Is it okay to run the LRA?

1. Yes, because none of the assumptions are violated, and the sample size is not important.

2. Yes, because the sample is big enough.

3. No, because there is multicollinearity.

4. No, because the dependent variable must be on the interval level.

Question 5

In a study with 80 subjects (40 females, 40 males) the odds that a subject seeks psychological help are 0.5. What does this mean?

1. The probability that a subject seeks help is twice as large as the probability that a subject does not seek help.

2. The probability that a subject seeks help is the same for females and males.

3. The probability that a subject does not seek help is twice as large as the probability that a subject seeks help.

4. The probability that a subject seeks help is equal to the probability that a subject does not seek help.

Question 6

With logistic regression there is only a relationship between X and ...

1. The odds.

2. The probability of belonging to group 1.

3. The log(odds).

4. The odds ratio.

Question 7

Which statement is true?

I. With dichotomous variables the presence of a trait/variable is indicated with 1, and the absence of it is indicated with 0.
II. LRA tries to predict to which group an individual belongs by calculating the probability that the individual belongs to the target group.

1. I is true.

2. II is true.

3. Both I and II are true.

4. Neither I or II is true.

Question 8

What is not a reason that the least-squares method is unsuitable for LRA?

1. The dependent variable is dichotomous, but must be of the interval level at the least.

2. The assumption of equal variances is almost never met.

3. The least-squares method generates values greater than 1 and smaller than 0.

Question 9

What statement is true?

1. Transformation occurs when using natural log, so that the data fits on the S-curve.

2. When the predicted probability is greater than .5, than the individual belongs to the target group (and gets the code 1).

3. The odds ratio is the odds of belonging to the reference group divided by the odds of belonging to the target group.

Question 10

Why is the least squares method named like this?

1. Because the lowest sum of squares gives an indication for the prediction of a certain variable.

2. Because the regression line is placed where the sum of the squared distances is as little as possible.

3. Because the lowest error of the sums of squares indicates that there is a better fit in the model.

Question 11

With which evaluation test do we test the null hypothesis that all coefficients are 0?

1. 2LL test.

2. Omnibus test.

3. Pseudo R2-test.

4. Hosmer and Lemeshow test.

5. Wald test.

Question 12

How do we interpret a significant Omnibus test?

1. The variances of the predictors aren't equally distributed.

2. The set of variables predicts the outcome variable significantly.

3. Dependent on the value, a statement can be made about the explained variance.

Question 13

An educational psychologists investigates if passing an exam can be predicted by the number of study hours with a LRA. This yields the following output:

Variables in the Equation

For which number of study hours is the probability of passing the exam equally as big as the probability of failing?

1. 11.5.

2. 46.

3. 48.

4. 50.

Answer indication

1. ​​​​C. The probability of passing becomes five times as large if you attend one more lecture.

2. D. .04.

3. D. A: 66, B: 64.3.

4. C. No, because there is multicollinearity.

5. C. The probability that a subject seeks help is the same for females and males.

6. C. The log(odds).

7. C. Both I and II are true.

8. A. The dependent variable is dichotomous, but must be of the interval level at the least.

9. B. When the predicted probability is greater than .5, than the individual belongs to the target group (and gets the code 1).

10. B. Because the regression line is placed where the sum of the squared distances is as little as possible.

11. B. Omnibus test.

12. C. Dependent on the value, a statement can be made about the explained variance.

13. B. 46.

How does mediation analysis work? - ExamTests 7

Questions

Question 1

What is true about (semi-)partial correlations?

1. Partial correlations indicate the percentage of unique variance explained by the variable.

2. The squared semi-partial correlation of X₁ is calculated by dividing the unique variance explained by the total variance of the dependent variable.

3. The squared partial correlation of X₁ indicates the percentage of variance that partly explains the dependent variable.

4. The squared multiple correlation (R²) is equal to all unique variance explained in total.

Question 2

A regression coefficient that explains the relationship between independent variable X and dependent variable Y is strengthened when we add the mediation variable M. What situation do we have?

1. A perfect/complete mediation.

2. A partial mediation.

3. The lack of mediation.

4. A suppression effect.

Question 3

What is not an assumption that must be met in mediation analysis?

1. The independent variable must significantly predict the dependent variable.

2. The error of the independent variable must disappear when the mediation variable is added to the model.

3. The independent variable must significantly predict the mediation variable.

4. The mediation variable must significantly predict the dependent variable.

Question 4

To analyse a mediation we use three regressions. Below you can see a table of that. Self-control is predicted by positive affect as the independent variable and self-confidence as the mediation variable. What statement is not true?

1. By adding self-confidence the model explains more variance.

2. Because all relationships are significant, we can assume that there is a mediation effect.

3. Adding self-confidence was useless, because the t-value was higher in path c; the unmediated scenario.

4. Positive affect and self-confidence significantly predict self-control.

Question 5

What is the situation in the same analysis?

1. Perfect/complete mediation.

2. Partial mediation.

3. A lack of mediation.

4. Suppression effect.

Question 6

What is the relative strength of the effect?

1. .554.

2. .402.

3. .527.

4. .473.

Question 7

A consumer psychologist investigates whether the number of read books can be predicted on the basis of reading pleasure. The psychologist suspects that this relationship is mediated by the number of books that someone buys on average per month. The psychologist performs a mediation analysis. This yields the following output:

Coefficients

Dependent Variable: Written books

Coefficients

Dependent Variable: Written books

Coefficients

Is there, according to this causal-steps approach of Baron & Kenny, a matter of no, or a partial, or a complete mediation, or of suppression?

1. No mediation.

2. Partial mediation.

3. Complete mediation.

4. Suppression.

Answer indication

1. B. The squared semi-partial correlation of X₁ is calculated by dividing the unique variance explained by the total variance of the dependent variable.

2. D. A suppression effect.

3. B. The error of the independent variable must disappear when the mediation variable is added to the model.

4. C. Adding self-confidence was useless, because the t-value was higher in path c; the unmediated scenario.

5. B. Partial mediation.

6. D. .473.

7. A. No mediation.