Correlational Statistics and Tests of Reliability

Question 1

Correlation

Answer 1

Used to test the relationship between variables (quantitative or categorical), how things are related. Can make predictions about future events.

Question 2

Reading Correlation

Answer 2

Have a visual notion about the relationship of variables using a scatterplot.
Weight vs height -: high correlation -> if one increases the other increases aswell.

Question 3

Correlation Coefficient (Pearson’s Coefficient)

Answer 3

Describes strength and direction of linear association between 2 continuous (interval or ratio) variables.
r= -1 -> Negative strong correlation (if one increases the other decreases)
r= 1 -> Positive strong correlation
r= 0 -> No correlation

Question 4

Spearman’s Correlation Coefficient

Answer 4

Used when you have qualitative ordinal/nominal scale variables or a mix of quantitative and qualitative.

Question 5

Coefficient of Determination

Answer 5

r2 (í öðru veldi). Ratio of amount of variance explained by the regression model to the total variation in the data.

Question 6

Reliability Tests

Answer 6

Consistency of a measure.
High reliability if a measure produces similar results under consistent conditions.

Question 7

Percent Agreement or K-statistics (Cohen’s K)

Answer 7

Most used with categorical variables. Determines how well an observation produces the same value, for the same patient, on repeated measurement.

Question 8

Percentage Agreement Formula

Answer 8

Sum of the agreed observations divided by total nr of observations.

Question 9

Crosstabulation - Observed, Expected, Reliability

Answer 9

Shows in tabular format the relationship between 2 or more categorical variables.
Observed: actual nr of cases within each cell.
Expected: Expected value for each cell.
Reliability: Relationship between what we observe and what we expect.

Question 10

Cohen’s Kappa Statistics Table

Answer 10

K Lvl of agreement
0-.20 None
.21-.39 minimal
.40-.59 Weak
.60-.79 Moderate
.80-.90 Strong
>.90 Almost perfect
K=0 amount of agreement expected from random chance.
Quantifies agreement beyond chance for categorical variables.

Question 11

Cohen’s Kappa Formula

Answer 11

K = Po-Pe/1-Pe
Po: Percent agreement observed
Pe: Percent agreement expected
Po=nr of observations agreed on from both categories/total nr of observations
Pe=(marginals of category 1 from rater 1 x marginals of category 1 from rater 2/total nr of observations) + (marginals of category 2 from rater 1 x marginals of category 2 from rater 2/total nr of observations)/total nr of observations.

Question 12

Coefficient of Variation - Description + Formula

Answer 12

Determines the relationship between standard deviation and the mean of two sets of observations (2 goniometric measurements). For continuous variables.
CV= (standard deviation/mean) x 100
Values closer to 0 show minimal variance.

Question 13

Interclass Correlation Coefficient (ICC)

Answer 13

Reliability measure to use in continuous variables. Between 0-1, always associated to 95% confidence interval. The higher the ICC the higher the reliability.

Question 14

Standard Error of Measurement (SEM)

Answer 14

Estimation of the expected random variation in scores when no real change has taken place. If the new measurement is higher than the SEM, it’s a real measurement. Everything below SEM is considered random chance.

Question 15

Minimal Detectable Difference (MDD)

Answer 15

Minimal amount of change that needs to be observed, at either the group or individual level, for it to be considered a real change. Take results and minus the SEM. If it’s higher than the MDD, it’s clinically significant.