Ch24Research

Question 0

Why is the Pearson product moment correlation alone not a complete tool for documenting reliability?

Answer 0

it assesses association (relative reliability) not concordance (absolute reliability); we want to know both the relationship between the two measures and the magnitude of the differences!

Question 1

What are three techniques used for reliability analysis? (relative reliability)

Answer 1

Pearson product moment correlation with regression and difference analysis extensions, intraclass correlation coefficients, kappa correlations

Question 2

What three strategies are used by researchers to supplement the information gained from the Pearson correlation coefficient?

Answer 2

paired t-test, slope and intercept documentation, and determination of the standard error of measurement (SEM)

Question 3

Would researchers typically use more than one extension for the Pearson measure of relative reliability?

Answer 3

not typically, any one of the extensions (paired t-test, regression equation with documented slope and intercept, or reporting the SEM) would be sufficient

Question 4

What are the intraclass correlation coefficients (ICC)?

Answer 4

a family of coefficients that allow comparison of two or more repeated measures, this technique depends on repeated measures ANOVA

Question 5

Why is the ICC thought to be a better measure than the Pearson r?

Answer 5

it accounts for absolute as well as relative reliability; researchers should still report the results of an absolute reliability indicator such as the SEM or repeated measures ANOVA when reporting reliability based on the ICC

Question 6

When three or more raters are used to measure each participant, which must be used: Pearson or ICC?

Answer 6

ICC, when only two raters are present the researchers may choose between Pearson and ICC

Question 7

What is kappa?

Answer 7

reliability coefficient designed for use with nominal data, adjusts the agreement percentage to account for chance agreements and can be weighted to account for the seriousness of the discrepancy

Question 8

What are multiple regression techniques?

Answer 8

designed to analyze complex relationships among many variables, traditionally they use numerical IVs to predict a numerical DV

Question 9

Can a multiple regression technique accommodate nominal IVs?

Answer 9

yes, if a different number is assigned to each of the nominal levels; ex. yes = 1 and no = 0

Question 10

What is used to designate the correlation between all the IVs and the DV in a multiple regression equation?

Answer 10

R (to distinguish it from r)

Question 11

Explain variable entry in multiple regression.

Answer 11

researchers often specify various decision rules to guide the computer in generating the regression equation, the rules are constructed so that the method of variable entry maximizes the accuracy of predictions while minimizing the number of variables in the equation (variables are retained if they improve R2)

Question 12

Compare forward and backward regression strategy.

Answer 12

forward - researcher adds one variable at a time and stops when the additional variables do not contribute to the preset amount; backward - researcher begins with all the possible variables of interest in the equation and deletes them one at a time if their presence does not contribute to the preset amount (stepwise regression combines the two)

Question 13

How can we assess the meaningfulness of the multiple regression equation?

Answer 13

F test of R2 can determine if it was significantly different from 0, generate a confidence interval, or determine the relative contribution of each of the variables to the equation through a t-test or dividing R2 into components attributable to each variable

Question 14

What factors influence the validity of multiple regression procedures?

Answer 14

the extent of multicollinearity (high levels of correlation among the predictor variables), cross-validation of the regression equation with a second sample of participants or with random subsets of the original sample, adequate sample size, or how outlying data points are handled

Question 15

Why would we use logistic regression?

Answer 15

to predict dichotomous outcomes

Question 16

What is the rationale for logistic regression?

Answer 16

S-shaped curve is typical of the relationship between a dichotomous variable and a continuous variable, the relationship is non-linear so linear correlation and regression techniques are not appropriate (logarithmic and exponential transformations used instead)

Question 17

What are the two common ways of presenting the results of logistic regressions?

Answer 17

odds ratios for variables within the logistic regression equation (individuals with visual problems were 2.80 times as likely to have multiple falls) or presenting the equation itself (individual without a history of imbalance and a Berg Balance Scale score of 54 would result in a prediction of a 5% probability of falling)

Question 18

What is discriminant analysis?

Answer 18

procedure in which multiple predictor variables are used to place individuals into groups, backward MANOVA procedure; ex. using set of combined communication variables from each participant to predict whether each is a boy or girl

Question 19

What is a factor analysis?

Answer 19

tool whereby correlational techniques are used to discover which of many variables cluster together as a related unit, separate from other, unrelated clusters; data reduction technique

Question 20

What are three reasons a factor analysis would typically be done?

Answer 20

test development (reduce a great number of items into a smaller number), theory development (examine underlying structure of a set of variables without a theoretical framework yet), or theory testing (items thought to be representative of certain constructs are factor analyzed to determine whether the items load as hypothesized)

Question 21

Give an example of how the factor analysis might be presented in the literature?

Answer 21

one study used factor analysis to determine what factors might underlie a wide range of hand evaluation items; the four factors identified were “pinch,” “grasp,” “target accuracy,” and “activities of daily living;” this set of factors could be used as the basis for developing a battery of hand function tests that would be more comprehensive than existing tests, none of which include items from ALL FOUR factors