Correlational Designs - Chapter 9
Terms in this set (40)
What are correlational designs?
A relationship between two or more variables; measured, not manipulated.
What does correlational research allow us to do?
Make predictions, examine the relationship between 2 variables
Each point on a scatterplot represents an individual score. A positive correlation starts low on the x and y axis and increases diagonally. A negative correlation starts high on the x and y axis and decreases diagonally. Zero correlation, the dots are randomly distributed.
Pearson's r coefficient of correlation
Indicates the strength of a correlation. Ranges from 1.00 (perfect positive) to -1.00 (perfect negative), with 0 indicating no correlation. It is a descriptive statistic.
What scales is Pearson's r used for?
Interval and ratio scales
What determines strength of correlation?
The relationship between the variables.
What assumptions are made about correlations with the Pearson's r?
It is assumed that the relationship is linear.
How does the Yerkes-Dodson Law relate to correlational research problems?
It is a curved line that does indicate a correlation, but because it is not linear, the Pearson's r would say that there is no correlation. Because one side is positive and one side is negative, it is canceled out and seen as though there is no correlation.
What range of values are used in correlations?
+1.00 to -1.00.
Is there a risk if restricted ranges of values are used? Why?
Yes. If a restricted range of values is used, the correlation will drop significantly because the number of values is not as high.
Coefficient of Determination (r2)
How much of the variability in one variable can be accounted for by a 2nd variable. Calculated by squaring the Pearson's r.
Calculate the "best fit" for correlated data and use the regression line to make predictions. Should only be conducted if a significant correlation exists between two variables.
Minimize the distance between the best-fit line and each point. Y = a + bX where a= y intercept, b= slope, X= known value (predictor variable) and Y = value you are trying to predict (criterion variable)
Can causal relationships be determined between the measured variables?
Correlational research does not imply causality, however, with cross-lagged panel correlations, we can begin to understand it.
There may be a correlation, but cause is not known. A and B could be correlated, but we do not know if A is causing B or B is causing A. We cannot infer causality from correlational research.
Cross-Lagged Panel Correlation
Helps solve the directionality problem. Investigates correlations between variables at several points in time (a type of longitudinal design), adding the causal element of A preceding B.
Third Variable Problem
Because correlational research does not necessarily control for outside variables, an outside variable can sometimes explain the correlation found. i.e., causes it.
Grandjean et al. (1998)
PCB and mercury intoxication on cognitive abilities in children. Third Variable Problem:
Beyond 2 variables (bivariate). Allows to analyze multiple factors and look for correlations between them.
Multivariate technique. Has one criterion variable and at least 2 predictor variables. Allows you to examine whether they predict as well as how strongly they predict
A multivariate technique. Takes many variables and correlates them with each other. Then, groups of the variables can cluster together and form factors.
Where did correlational research originate?
In 1888 when Sir Francis Galton published "Co-relations and their Measurement", where he looked at the heights of parents and children
What was significant about Lee Cronbach's 1957 "two disciplines" address?
Cronbach was worried that correlational research was seen as secondary to experimental and called for them to be valued equally, and perhaps even united.
Positive vs negative bivariate correlations
In positive, a high score on one variable is associated with high score on the other (or low score on one, low score on the other). In negative, it is an inverse relationship- if one is high, the other is low. If one is low, the other is high.
What can influence the size of correlation coefficients?
The strength of the correlation- the stronger the correlation, the closer it is to 1.00 or -1.00
How does regression analysis help us to predict?
By using a regression line, which summarizes the points on the scatterplot in the best way. This allows us to make predictions.
Criterion vs predictor variables
Criterion variables are usually the Y value, which is what is trying to be predicted. Predictor variables are usually the X value, whose value the prediction is based on. i.e., time spent studying on GPA. Predictor would be time spent studying and criterion would be GPA.
How can directionality make it difficult to interpret correlations?
Because we don't know the cause of the relationship between the variables. We do not know whether A causes B or B causes A.
How does a cross-lagged panel design help with the directionality problem?
It helps by studying the variables at different points in time, adding the causal element.
How can the third variable problem be evaluated and controlled?
Partial correlation can control for third variables. Takes into account the third variable by taking it out of the correlation and seeing the remaining relationship between the variables that were originally studied.
In what research situations are correlational procedures likely to be used?
When studies cannot be done for physical or ethical reasons, to study differences in personality and in psychopathology, and in twin studies looking at the heredity of a trait.
Multivariate procedures of multiple regression and factor analysis
Multiple regression has 1 criterion variable and at least 2 predictor variables that are tested for it. Factor analysis has many variables that are all tested for correlation with each other.
If a Pearson's r is +.88, what is the coefficient of determination, and what does it mean?
The coefficient of determination is .7744 and it implies a strong correlation.
What is split-half reliability?
A procedure used to evaluate the reliability of a test. Involves splitting the items that make up a subtest in half, and correlating the 2 halves. If it is reliable, the correlation should be high.
When assessing split-half reliability, what are the two sets of numbers being correlated?
Each set of numbers is made up of half of a subtest (i.e. even numbers or odd numbers)
How does bivariate regression differ from multiple regression?
In bivariate regression there is only one predictor variable. In multiple regression, there is more than 1 predictor variable. However, they both only have 1 criterion variable
How does multiple regression differ from factor analysis? What do they have in common?
In multiple regression, 2+ variables combine to predict an outcome. In factor analysis, the goal is to identify groups of variables that correlate highly with each other. Both are multivariate techniques and both involve the measurement of more than two variables
If you wish to predict academic success in college by looking at high school grades, what is the criterion variable and what is the predictor variable?
College GPA would be the criterion variable and high school grades would be the predictor variable.
How does the directionality problem make it difficult to interpret a correlation between depression and lack of exercise?
We do not know if the depression causes the lack of exercise, or if the lack of exercise causes depression.
In the Eron et al. (1972) study on aggression, which techniques were used to address a) the directionality problem and b) the third variable problem?
A cross-lagged panel correlation and partial correlation