Test Questions Flashcards
(108 cards)
1
Q
The linear regression model is also called the OLS regression because:
A
It is based on the smallest sum of squared errors between the regression line and the raw data.
2
Q
In logistic regression, to see if the predictor in the regression model is better than the y-intercept only in the regression model, the:
A
Null deviance should be greater than the Residual deviance.
3
Q
In a randomized control experiment for severe depression, there are 2 groups: psychotherapy versus waitlist control. Those in the waitlist control group got less depressed from baseline to the end of the treatment period. One explanation for this is:
A
Regression to the mean.
4
Q
The logistic regression uses a statistical method to create a predictive model that makes the sample data most likely to occur or be predicted. This method is called the:
A
Maximum likelihood method.
5
Q
Random intercept
A
Paired with coefficient (u0), it represents cluster-level deviations from the population average.
6
Q
Fixed Slope
A
The population average of the effect of the predictor on Y-hat.
7
Q
Random Slope
A
Paired with coefficient (u1) it allows for deviations from the population average.
8
Q
Fixed Intercept
A
The population average of the y-intercept effect.
9
Q
Table: Simple slopes of PHQ and gaming on post-gaming mood.
What is the outcome variable?
A
Post-gaming mood.
10
Q
Table: Simple slopes of PHQ and gaming on post-gaming mood.
How would you define the depression variable as used in this study?
A
A continuous moderator.
11
Q
Level
A
A component of a data hierarchy. Level 1 is the lowest level, for example, students within schools or repeated measurement occasions within individual subjects.
12
Q
Cluster
A
A grouping containing ‘lower level’ elements. For example, in a sample survey the set of households in the neighborhood.
13
Q
Nesting
A
The clustering of units into a hierarchy.
14
Q
Random part
A
That part of a model that is the contribution of the random variables at each level.
15
Q
Change in depression will be significantly greater for those in the CBT group compared to the WLC group
A
OLS Regression with a dummy code for treatment. WLC is the reference group.
16
Q
Either treatment (i.e., CBT or DBT) increase the odds of finding a job after leaving the inpatient care center compared to WLC.
A
Logistic regression with a WLC reference group.
17
Q
Depression will vary across the 10 clinics, and rate of change will likely vary by clinics across levels of treatment (CDT, DBT, WLC) with respect to depression at time 1 vs. time 3.
A
Mixed effect model with a random intercept, WLC is the reference group for the random slope of treatment.
18
Q
Quality of life based on a 10-question measure with a 7-point likert scale.
A
Interval
19
Q
Number of days
A
Ratio
20
Q
Type of anxiety disorder (GAD, Social anxiety, Specific phobia)
A
Nominal
21
Q
Symptom severity (mild, moderate, severe)
A
Ordinal
22
Q
Fixed Effects
A
This reflects the group average y-intercept and regression coefficient (slope).
23
Q
Level 2 (Person) Random Effects
A
This reflects predictors based on individual y-intercepts and slopes.
24
Q
Level 1 (Occasion) Random Effects
A
This reflects the error or deviation of the individual subjects’ raw scores from the predicted y-scores.
25
In MEM, to determine if studying alone will lead to more improved test scores over time than studying in small groups, which effect or term should be significant?
The interaction of study type and time must be significant.
26
When doing group comparisons for each time point of a longitudinal study, the research must worry about:
Experimentwise Type I error.
27
Order from first to last:
- Center variables
- Check for interaction
- Run multiple regression
- Interpret post-hoc tests
- Center variables
- Run multiple regression
- Check for interaction
- Interpret post-hoc tests
28
"Results of the model indicated that the odds of a patient who is treated being cured are 3.42 times higher than those of a patient who is not treated." This result is an example of:
Logistic regression.
29
Correlation
Fully standardized
30
Covariance
Scale dependent
31
Regression coefficient
Partially standardized
32
Correlation
Scale-invariant
33
Covariance
Unstandardized
34
Not fully standardized
- For every three inches in height, openness increase .25 likert scale points.
- Covariance.
- b-weight
- For every 1 unit increase in depression, there is a corresponding 4 unit increase in depression as measured by the BDI.
35
Fully standardized
- Correlation
- There is a small correlation between X & Y, such that as X increases, Y decreases (r = .18).
- For every standard deviation increase in happiness there is a corresponding .2 standard deviation increase in quality of life.
- beta-weight
36
SS Regression
variability in Y that can be explained by the predictor, represents the component Y that is shared with X1
37
SS Residual
variability in Y that cannot be explained by the predictor, simply what is 'left over' after accounting for x
38
Imagine we are looking at depression scores (y) and have IVs of age (x1) and distance from the therapist's office (x2). We want to evaluate the main effects for age & distance using regression. We wisely decide to center the variables.
Our centered coefficient (bo) is the value of height when the age is _______ and distance from the office is ______.
Average; average.
39
In multiple regression with continuous independent variables, we need to center the independent variables around the means before we center the interaction variable because...
We want the correlations between the independent variables & interaction to be zero.
40
In multiple regression, if the correlation between X2 and Y is zero, but when X2 entered into the multiple regression in Step 2, it causes the R squared to increase, this is called:
Supression
41
A partial correlation will always be larger than the semi-partial correlation because...
The denominator of the partial correlation is smaller.
42
In dummy coding for a multiple regression, b1 is...
The difference between the mean of the reference group and the mean of another treatment group.
43
R output.
Look at adjusted R-square value to determine stress predicting depression severity.
44
In multiple regression, there are different ways to enter the predictors (independent variables) in the statistical model. Which one involves entering all of the predictors at the same time?
Simultaneous Entry
45
The b-weight indicates difference between conditions.
Dummy coding
46
The b-weight is used to calculate difference between conditions.
Contrast Coding
47
The code for two groups is 0, 1.
Dummy Coding
48
The code for two groups is 1, -1.
Contrast Coding
49
One group is a reference group. Best for when 1 group is a control group and the other groups are to be compared to that group.
Dummy coding
50
Used when there are apriori hypotheses about the specific pattern of differences about the group means.
Contrast coding
51
The code for two groups is 1, -1 in the case of the unweighted type, we consider the proportion of cases in each group.
Effect coding
52
The b-weight indicates difference between the group mean & the grand mean.
Effect Coding
53
Used when comparing 1 group's mean to the total sample's (weighted or unweighted) mean.
Effect Coding
54
The linear regression model is also called the OLS regression because...
It is based on the smallest sum of squared errors between the regression line and the raw data.
55
In logistic regression, to see if the predictor in the regression model is better than the y-intercept only in the regression model the...
Null deviance should be greater than the Residual deviance.
56
In a randomized control experiment for severe depression, there are 2 groups: psychotherapy versus waitlist control. Those in the waitlist control group got less depressed from baseline to the end of the treatment period. One explanation for this is:
Regression to the mean
57
The logistic regression uses a statistical method to create a predictive model that makes the sample data most likely to occur or be predicted. This method is called the...
Maximum likelihood method
58
Paired with coefficient (u1) it allows for deviations from the population average.
Random slope
59
Paired with coefficient (u0), it represent cluster-level deviations from the population average.
Random Intercept
60
The population average of the y-intercept effect.
Fixed Intercept
61
The population average of the effect of the predictor Y-hat.
Fixed slope
62
How many levels?
Time in students (subjects) in classrooms in school in districts.
5 levels.
63
In MEM as a regression to make interaction variables with a continuous predictor (independent variable), what must you first do to the continuous predictor?
Center the scores around the mean.
64
To interpret the fixed effects of the main effects of independent variables and their interaction, you should look at the...
Type III Tests of Fixed Effects Table
65
In a MEM regression, how many vectors do I need for a continuous predictor (IV) called anxiety severity?
1
66
In an MEM regression, how many vectors do I need for a nominal predictor with 4 treatment groups?
3
67
The problem with General Linear (ordinary least squared) Model for longitudinal studies is that if a subject is missing data for 1 time point...
You cannot use any of the data for that subject.
68
A logistic regression was conducted to see if having kids (kid_dum) can predict missing data. If having kids did predict missing data, then you should _________. This type of missing data is consider ____.
Include kid_dum in the MEM model; MAR.
69
Individuals lower in working memory capacity are particular vulnerable to anxiety's disruptive effect on performance.
The independent variable (s) is/are:
WM capacity and Trait Anxiety
70
Individuals lower in working memory capacity are particular vulnerable to anxiety's disruptive effect on performance.
The primary threat to internal validity in the present design is:
P's are not randomly assigned to different levels of the IV(s)
71
The present of a significant interaction suggests:
The researcher should not interpret the main effects, and instead focus on conditional or "simple effects"
72
Cohort effects can negatively impact experimental validity because:
Unique characteristics between the groups may influence the results of the study.
73
The form of reliability based on the same participants' response to a measurement instrument administered at two different time points is called:
Test-retest reliability
74
A researcher looks at the correlations between a depression measure, a self-esteem measure, and a math skills measure. he theorizes that depression is conceptually related to self-esteem, but not math ability. if the correlation between depression and self-esteem is high that would provide evidence of ________. If the correlation between depression and math ability is low, that would provide evidence of _______.
Convergent validity, Discriminant validity.
75
The generalizability of experimental results is most dependent upon:
The representativeness of the sample participants.
76
Why is variability in the data at baseline assessment problematic in single-case designs?
Because performance at baseline is used to predict future performance and variability makes prediction difficult.
77
What steps are typically taken to reduce experimenter expectancy effects?
Keeping the experimenter blind to the condition.
78
Researchers who conduct research with individuals having limited ability to understand the nature of the research and the possible risks:
Must obtain informed consent from the parents or legal guardians and should strive, when feasible, to get assent from the participants themselves.
79
A neuropsychologist did an intensive study of a person who had suffered an unusual brain injury that resulted in the better understanding of types of memory loss and the brain. Which of the following methods did the psychologist use?
Case study method.
80
A researcher who argues that the independent and/or dependent variables were manipulated or measured inappropriately in a study is criticizing the _____ validity.
Construct
81
Presenting substantial portions of another person's written work as your own is called...
Plagiarism.
82
The Census Bureau collects basic demographic data for all persons residing in the United States. If we obtain census data for 2010 and compute the frequency of female children (age < 18), we will obtain a _____, because our computation is based on ______ data.
Parameter; population.
83
The number of affiliative behaviors for a group of n = 6 children were recorded over the course of one hour: 3, 10, 3, 4, 3, 7. In this distribution of scores, the mean is ____ the mode and _____ the median.
Greater than; greater than.
84
If the shape of a frequency distribution is asymmetrical, with the tail projecting longer into the left than to the right, how would the distribution be skewed?
Negatively
85
Incorrectly rejecting Ho when there are actually no differences at the population level is known as _____ error.
Type I
86
A distribution of sample variances based on random samples of a fixed size from a population is a ________.
Sampling distribution
87
The formula that used to compute the estimated standard error of a sample mean is based on _________.
Central Limit Theorem
88
For a test of mean differences (i.e., single sample vs. population; independent samples; dependent samples), we use the _____ when the population variance is known, and the ____ when the population variance is estimated using the sample data.
Z distribution; t distribution
89
For an independent-groups design (i.e., Tx vs. Ctrl), effect size can be measured using...
Cohen's d or Hedge's g
90
If a single-sample t-test and a dependent samples t-test both have df = 16, which study used more participants?
They both used the same number of participants (n =17)
91
In the context of ANOVA, which situation is most likely to produce a large value for the F-ratio?
large group-mean differences and small within-group sample variance
92
If you have a 3 x 3 between subjects design and run a total of 220 subjects, the degree of freedom in the error term will be:
211
93
In a 3 x 3 x 3 between subjects design, how many vectors will be needed to produce the main effects and the interactions in the source table?
26
94
Which of the following is an assumption of multiple regression?
Normal distribution of errors (residuals), linearity, homoscedasticity
95
A statistical model examining how the effect (i.e., slope) of depression (IV) on memory performance (DV) changes depending on a person's age describes what phenomenon:
Moderation
96
A statistical model examining how the effect (i.e., slope) of depression (IV) on memory performance (DV) changes depending on a person's age describes what phenomenon:
Moderation
97
A researcher would like to compare one (reference) group's mean to each of the other groups (individually) using multiple regression. Which coding system should this person use?
Dummy coding
98
The covariance between these two variables describes a _____ relationship expressed in ______ scale.
linear; their original (raw)
99
The following multiple regression equation expresses depression scores as a function of the number of prior major depressive episodes, adverse childhood experiences, and the interaction: Dep = bo + b1 * PriorMDE + b2 * ACE + b3 * PriorMDE x ACE + e dep. Assume we run the analysis twice, the first time with predictors uncentered, and a second time with the predictors centered around the grand mean. Which parameters will differ across these two analyses?
bo, b1, and b2
100
Regression to the Mean (RtM) describes the process by which ____ extreme observations at Time 1 become _____ extreme at Time 2.
more; less
101
Scores on a variable with a Time 1 - Time 2 correlation (RT1-T2) near 1 (or -1) will show ____ regression to the mean than a variable with rT1-T2 near 0.
less
102
In the context of Mixed Effects Modeling (MEM) ___ effects describe means and/or regression coefficients that are constant across all individuals in a group, whereas ____ effects describe person- or time-specific deviations.
fixed; random
103
Is possible to distinguish between data that are ______ and ______.
MAR and MCAR
104
When analyzing longitudinal data with missing responses, the Full-Information Maximum Likelihood (FIML) estimator used in Mixed Effects Modeling (MEM) provides more accurate results (relative to traditional OLS ANOVA using listwise deletion) when data are _______.
MAR or MNAR
105
For two variables (X, Y), the SP = +18. The covariance between these two variables will be....
Positive.
106
The covariance between these two variables describe a _____ relationship expressed in _____ scale.
linear; a standardized
107
A negative value for a correlation indicates _______.
Decreases in X tend to be accompanied by increases in Y.
108
A college professor reports that students who finish exams early tend to get better grades than students who hold on to exams until the last possible moment. The correlation between exam score and amount of time spent on the exam is an example of ______.
A positive correlation.
