Deck 21: Multilevel Linear Models

Question 1

Prior to conducting a multilevel analysis, how might you check linearity and any unusual cases (e.g. outlying cases)?

A) Scatterplots
B) Bar charts
C) Pie charts
D) Radar plot

Answer 1

Scatterplots

Question 2

Which of the following is not a form of polynomial curve?

A) Linear
B) Quadratic
C) Cubic
D) Binomial

Answer 2

Binomial

Question 3

Based on a multilevel analysis, a chi-square value of 7.05 was obtained at one degree of freedom. What could be concluded from this finding?

A) The change is not significant.
B) The change is significant at the .10 level.
C) The change is significant at the .05 level.
D) The change is significant at the .01 level.

Answer 3

The change is significant at the .01 level.

Question 4

What is the purpose of multilevel linear analysis?

A) To analyse data with a hierarchial structure.
B) To analyse data from independent groups.
C) To analyse data that are logarithmic.
D) To analyse data that are not normally distributed and do not have the assumption of sphericity.

Question 5

The UK coaching pyramids in rugby essentially have five levels, with level 1 introducing individuals to coaching, whilst level 5 is tailored for those working at professional clubs or international teams. How would you describe the effects if all five levels were analysed?

A) Fixed effects
B) Random effects
C) Variable effects
D) Intercept effects

Question 6

Which of the following descriptions best applies to this graph?

A) Fixed intercept, random slope
B) Fixed intercept, fixed slope
C) Random intercept, random slope
D) Random intercept, fixed slope

Question 7

Where would you position the club (i.e. what level) in the model of Q3?

A) 3
B) 4
C) 5
D) 6

Question 8

A researcher wanted to investigate the confidence levels of athletes in different athletics clubs. In addition, he wanted to see whether the athletes' coaches had an influence and whether this varied from country to country. How many levels would be involved in this analysis?

A) 3
B) 4
C) 5
D) 6

Question 9

Sports and exercise science related research frequently uses small sample groups. Therefore, when applying a multilevel analysis technique it is necessary to calculate the log-likelihood value. SPSS produces four versions of the calculation, but which of the following would be most appropriate to use when sample sizes are small?

A) Hurvich and Tsai's criterion (AICC)
B) Akaike's information criterion (AIC)
C) Bozdogan's criterion (CAIC)
D) Schwarz's Bayesian criterion (BIC)

Question 10

What is the minimum number of levels required to perform a multilevel linear analysis?

A) 2
B) 3
C) 4
D) 5

Question 11

Which of the following would not be expected to change with time?

A) Attitude towards training
B) Gender of participants
C) Anxiety prior to competition
D) Cellular mitochondria density during aerobic training

Question 12

Suppose you were interested in the effect of parental influence (x-axis) on a child's confidence (y-axis) when playing rugby. In addition, you examine the coach's qualifications . What will be the most likely appearance of the intercept and slope?

A) Fixed intercept, random slope
B) Fixed intercept, fixed slope
C) Random intercept, random slope
D) Random intercept, fixed slope

Question 13

Based on the example above, how many rows are required (in data view) for each individual surveyed in this study for a multilevel analysis to be conducted in SPSS?

A) 1 row per person
B) 2 rows per person
C) 3 rows per person
D) 4 rows per person

Question 14

Suppose you were interested in the development of player confidence when exposed to structured coaching sessions. What is the name of the model that examines the rate of development or change in a given variable (e.g. confidence)?

A) Growth model
B) Seismic model
C) Formulation model
D) Analysis model

Question 15

What is a second-order polynomial also known as?

A) Linear
B) Quadratic
C) Cubic
D) Binomial

Question 16

If a lack of linearity is identified as an issue, how might you address this?

A) Transform the data.
B) Change the test to a chi-square test.
C) Linearity is not important within multilevel analysis.
D) Conduct group mean centring.

Question 17

Which of the following descriptions best applies to this graph?

A) Fixed intercept, random slope
B) Fixed intercept, fixed slope
C) Random intercept, random slope
D) Random intercept, fixed slope

Question 18

One of the issues with multilevel linear analysis is that it introduces dependency between the contextualized variables (e.g. the influence of different coaches on their athletes' confidence), but what does this mean?

A) Residuals within the model will be correlated.
B) The variables are from different cases and therefore make the model unstable.
C) Independence of the different levels makes the analysis more prone to Type I error.
D) The experimental nature of the research design implies that the contextualized variables will differ at each level of the model.

Question 19

Group mean centring is a useful way to prevent multicollinearity, but how is it usually calculated?

A) An individual score (player confidence) is subtracted from a group mean (mean player confidence in that group).
B) An individual score (player confidence) is multiplied by the group mean (mean player confidence in that group).
C) An individual score (player confidence) is added to a group mean (mean player confidence in that group).
D) An individual score (player confidence) is divided by a group mean (mean player confidence in that group).

Question 20

A covariance structure needs to be applied to multilevel modelling when there are random effects or repeated measures incorporated into the design. These are important as they provide an estimate of the model parameters, but what happens if you apply a covariance structure that is too simple?

A) It increases the likelihood of making a Type I error.
B) It increases the likelihood of making a Type II error.
C) It increases the likelihood of making a Type I or Type II error.
D) It increases the likelihood of accepting the null hypothesis when it is false.