Statistics Final down to quiz 8

Among the following values of Pearson r, which indicates the strongest linear relationship between the two measured variables:

-0.80

Cohen has defined a small difference between means as one of _____ standard deviations.

.2

Cohen has defined a medium difference between means as one of _____ standard deviations.

.5

For Exercise 2, what is the value of d-diff (Cohen's d adjusted for the power-enhancing effect of the correlated samples design), aka "effect size dz." (Quiz 11)

.50

in independent samples t assume that rho =?

0

Each of 20 subjects' systolic blood pressure is tested twice -- once while relaxing comfortably and once after being insulted by another subject (not a real subject, an experimental confederate pretending to be a subject). We use a correlated t test to see if such an insult affects systolic blood pressure. What are the degrees of freedom?

19

You have data on degree of political conservatism for two groups of persons. One group consists of persons whose highest degree is a high school diploma. The other group consists of persons whose highest degree is a bachelor's or graduate degree. You are going to use a pooled independent samples t test to test the null hypothesis that mean political conservatism is the same in the two groups' populations. You have 10 scores for the one sample and 10 for the other and the variances are homogeneous. What is the critical absolute value of Student's t for a two-tailed test using the traditional 5% criterion of statistical significance?

2.101

Which of the following values is the critical absolute value of Student's t for alpha = .05 when degrees of freedom equal 6?

2.447

You are testing the difference in effectiveness of two types of therapy. The dependent variable is a measure of psychological wellness after the therapy. Subjects are matched up on pre-therapy wellness, and one member of each pair assigned to Therapy A, the other to the Therapy B. There are 16 patients assigned to A, 16 to B. The mean for A is 15 and for B is 10. The standard deviation of the difference scores is 8. What is the observed value of the correlated samples t?

2.500

In a bivariate regression analysis, the SSregression = 900 and the SSerror = 2100. What percentage of the variance in the dependent variable is explained by the linear model:

30

From the table below, what is the value of the intercept for predicting guilt from physical attractiveness

6.674

What is factoral ANOVA design?

A design where every level of every factor is paired with every level of every other factor

What is 2x3 factoral design?

A design with two independent variables, one with 2 levels and the other with 3 levels

Factor

Another term for independent variable used in factoral ANOVA

Which of the following APA-style summary statements does the best job of correctly presenting the results in proper APA style?

College students' quiz performance when tested on material previously presented while they were texting during lecture (M = 51.56, SD = 10.12) was significantly lower than when tested on material previously presented while they were not sending texts (M = 59.63, SD = 12.16), t(15) = 3.28, p = .005, d = .72.

What does MS(error) represent

Error variance

A Type I error is when the null hypothesis is actually false but you have not rejected it.

False

A distribution of sample means will have a mean and a standard deviation equal to the mean and the standard deviation of the population from which the samples were randomly drawn.

False

A robust statistical test is one which is greatly affected by violation of its assumptions

False

Both eta-squared and omega-squared estimate the proportion of variance explained by the ANOVA model, but eta-squared is a less biased estimator than is omega-squared.

False

Fisher's LSD is never an appropriate method for making pairwise comparisions, because it always allows the actual familywise alpha to exceed its nominal level.

False

For a fixed number of subjects, the independent t-test's power is less with equal sample sizes than with unequal sample sizes, ceteris paribus

False

Homoscedasticity means that error variance is correlated with X.

False

I am analyzing the relationship between a dichotomous X and a continuous Y. If I use an independent samples t-test and get significant results, I can conclude that variance in X is a cause of variance in Y, but if I use a t test of the null that the point biserial correlation between X and Y is zero and get significant results I cannot make any conclusion with respect to causality, I can only conclude that the two variables are related.

False

If Pearson r = 1, SSerror = SSy.

False

Increasing the sample size should make the width of a confidence interval (about the mean) greater.

False

Less than 2.5% of the area under a t-curve with 10 degrees of freedom is beyond (above) the value of t equals 1.96.

False

The absolute critical value for a two-tailed .05 t-test is generally lower than for a similar z-test.

False

The confidence interval for predicting the average value of Y given X is wider than that for predicting an individual value of Y given X.

False

The larger the effect size, the larger beta, ceteris paribus.

False

The within-subjects design is also known as the independent samples design

False

When conducting a one-sample t test, the degrees of freedom are equal to the sample size plus one, that is, N + 1.

False

With t-tests, the greater the number of degrees of freedom, the higher the critical value and the less the power.

False

Regression towards the mean is maximal when r equals -1.00.

False*

Which of the below multiple comparison techniques should be used only following a significant ANOVA:

Fisher's LSD

When would You likely use d family effect size measure in an analysis of variance?

Generally only when comparing 2 groups or sets of groups. Imprecise when comparing more than 2 groups at a time.

You have data on the severity of symptoms of anxiety in two groups of patients. One group has been treated with a traditional type of therapy. The other group has been treated with a new experimental therapy. You compute the strength of effect estimate by taking the difference between the two group means and then dividing that difference by the standard deviation of the group that got the traditional therapy. The name of this strength of effect estimate is:

Glass' delta

In a one-way independent samples equal n's ANOVA the average within group sample variance, [s21 + s22 + ... + s2k] / k is the:

MSE [Mean Square Error]

In an APA-style research report, which of the following symbols should be set in italic font?

N, M, and SD

What are two main assumptions behind ANOVA

Normality, homogeneity of variance, independence of observations

What does r family effect size communicate in factoral ANOVA?

Percentage of variance that can be accounted for by difference in the effect under consideration

How is the error of sum of squares calculated in factoral design?

Sum of squares left over after subtracting the main and simple effect sums of squares from the total sum of squares. It's variability does not belong to any of the effects in the design.

What is often the most important part of factoral ANOVA

The ability to examine the interaction between variables

What is a cell in factoral ANOVA?

The data of all participants at one level of one variable and one level of the second variable

What is simple effect

The effect of one variable at one level on another variable at one level

What does MS(between) represent?

The variability of group means

1 out of 1 points If the population standard deviation is known rather than estimated from the sample data, one uses z instead of t.

True

A robust statistical test is one which is not greatly affected by violation of its assumptions.

True

Anything that decreases the overlap of the null and alternate sampling distributions will increase the power of a two-tailed t-test

True

Even though X and Y are well correlated, it might be the case that X is not the cause of Y and Y is not the cause of X

True

Familywise alpha is always less than or equal to per comparison alpha times number of comparisons.

True

For a fixed number of subjects when conducting an independent samples t-test, power is greater with equal sample sizes than with unequal sample sizes, ceteris paribus.

True

For each of 20 married couples I have two scores, measures of political conservatism. For each couple I subtract the wife's score from the husband's. I calculate t as the mean of these difference scores divided by their standard error. I have calculated a matched pairs (correlated samples) t.

True

If the scatter plot indicates that the relationship is monotonic but not linear, computing the Spearman correlation coefficient is more appropriate than computing the Pearson correlation coefficient.

True

In ANOVA the expected value of F, if the null hypothesis is true, is approximately one.

True

One can accept a "loose" null hypothesis that e cannot reject if the analysis was very powerful.

True

One potential source of "error" in a linear regression analysis is the nonlinear "effect" of X upon Y.

True

Parametric tests, like the independent samples t test, often have assumptions about the shapes and homogeneity of variances of the populations compared.

True

Requiring that a preliminary ANOVA be significant prior to conducting pairwise comparisons among means with the REGWQ procedure is as foolish as wearing a condom and abstaining altogether from all sexual contact.

True

The obtained value of t, for a correlated samples t test, is 2.201, on 12 degrees of freedom. This is significant at the .05 level, two-tailed

True

The one-way ANOVA provides a test of the null hypothesis that each of k population means is equal to each of the other (k-1) population means

True

The related samples t test can be conducted by finding, for each pair of scores, the differences between the scores and then conducting a one-sample t test of the null that the mean difference is zero.

True

The within-subjects design is also known as the repeated measures design.

True

To establish that event X causes event Y, it is necessary to establish that X and Y are correlated

True

To estimate Cohen's d, we take the difference between two means and then divide by the standard deviation.

True

When pooling sums of squares one must assume homogeneity of variance.

True

A parametric null hypothesis always contains an equals sign (that is, the hypothesis that contains the equals sign is always the null hypothesis).

True*

Regression analyses on two different bivariate data sets (A & B) may reveal that they have identical slopes but different correlations

True*

What should you say about main effects if there is a significant interaction?

Usually ignore main effects although there is no rule that says to. Usually focus on simple effects.

You are testing the difference in effectiveness of two types of therapy. The dependent variable is a measure of psychological wellness after the therapy. Subjects are matched up on pre-therapy wellness, and one member of each pair assigned to Therapy A, the other to the Therapy B. There are 16 patients assigned to A, 16 to B. The mean for A is 15 and for B is 10. The standard error of the mean for the difference scores is 1.847. The two-tailed critical value of t for alpha = .05 is 2.131. What is the 95% confidence interval for the difference between means.

[1.064, 8.936]*

null hypotheses always or never include an equality sign?

always

cohen's d

effect size measure used in Student's t

With nondirectional hypotheses, the appropriate p value for ANOVA is one tailed, the probability of obtaining an F as large or larger than that computed, were the null hypothesis true.

false

esitmator for cohen's d in student's t

hedge's g (d hat)

t test with pooled variances is used when

homogeneity of variances can be assumed

If we obtain a negative r, this means that:

individuals scoring high on one variable tend to score low on the second variable

large rho means ....standard error

lower

this test is appropriate whenever data have been collected in a way that results in the two samples of scores being related to each other

matched pairs t

large rho means more or less power

more

matched pairs is ... than independent samples t because it has....

more powerful...smaller standard error

aletrnative hypotheses always or never include an equality sign

never

students t assumptions

normally distributed population, homogeneity of variance

Using the underlining method to summarize the results of pairwise comparisons, two means underlined by the same line are:

not significantly different from one another

Sol's IQ is 2 standard deviations above the mean. The correlation between IQ and GPA is r = + 0.50. When we predict Sol's GPA we predict that it will be ___ standard deviation(s) _____ the mean on GPA.

one; above

relationship where every time X goes up, Y goes up, as illustrated below, is a _____ relationship

perfect positive monotonic

in matched pairs rho is...

positive and fairly large

estimate p with ... because sample variances are likely unknown in independent samples t

r

matched pairs t is aka

related samples t, the correlated samples t, the within-subjects t, the repeated measures t, and the split-plot t.

Suzy Q measures each of 30 children's height Tuesday morning (scores = X) and does the same again Wednesday morning (scores = Y) with the same children. Then she computes the r on X,Y to estimate the _____ of her method of measuring height.

relliability

if homogeneith of variance cannot be assumed, use the .... to compute t

separate variances error term

Megan Waggy's thesis research demonstrated that employees' scores on a measure of organizational commitment were _________ with scores on a measure of intention to leave the organization sometime in the future.

significantly, negative correlated

Positively skewed data can be made less skewed by

taking the logarithm or square root of each score

The sum of squared deviations between actual Y scores and predicted Y scores is:

the SSerror

For the regression equation, Y = a + bX + error, "a" is:

the Y-intercept

homogeneity of variance

the assumption that two populations hav homogenous variances

mean

the average

p (rho)

the correlation between scores in population 1 and scores in population 2

median

the midpoint

Correlated samples t will have a smaller standard error (denominator of the t ratio) than independent samples t, leading to a larger absolute value of t and a better chance of obtaining a significant result if:

the scores in the one condition are well and positively correlated with those in the other condition

In an independent samples t test, the absolute value of t will be larger, and the null hypothesis more likely to be rejected, if:

the standard deviations in the two groups are small

pooled variances estimate

the weighted average of two sample variances

heterogeneity of variance may be removed by.... the data prior to analysis. this can be accomplished by...

transforming...square root and logarithmic transformations

Zimmerman says that separate variances t should be used

unconditionally whenever variances are unequal

One needs to assume that the population distribution of variable Y is normally distributed when one

uses t to test the null hypothesis that Y is not correlated with another variable