Chapter 12: Correlational Studies Flashcards
Purpose of a correlational study
to establish that a relationship EXISTS between the 2 variables, and to describe the nature of the relationship. IT DOES NOT MANIPULATE ANYTHING!
T/F A correlational study provides an explanation for the relationship
false.
how are correlational variables graphed?
via a scatter plot
What does the correlational coefficient describe?
describes 3 characteristics:
1) direction of the relationship (positive, negative)
2) format of relationship: linear exponential and inverse,
3) consistency or strength of the relationship– how scattery of the points?
Pearson coefficient:
used to describe and measure LINEAR relationships when both variables are NUMERICAL VALUES
T/F A pearson coefficient can be used for ordinal data
false. pearson coefficient can only be used when data is numbers
monotonic relationship
a relationship that is consistently one-directional, but the magnitude of increasing may not be the same size (ex/ michaelis menton graph)
What type of correlation can help describe a monotonic relationshiP?
a spearman coefficient
At what point does a correlational research study becoem differential?
when one variable is a non numerical variable and the other is a numerical variable, and when the NON NUMERICAL VARIABLE BECOMES THE VARIABLE THAT SEPARATES THE GROUP
ex/ gender and IQ levels. groups separated by male or female and IQs are measured. male and females are non numerical and thus this study is a differential reserach study.
what is a point biserial correlation?
when the non-numerical groups have only 2 categories and they can be assigned a 0 or 1 to determine a correlational coefficient.
T/F: the sign of the correlation is meaningless in a point biserial correlation
true, but the strength of the relationship between the non numerical factors can be determined.
What can you use as statistical analysis when determining the extent of a correlation between 2 non-numerical variables?
ex/ gender (male and female) vs success (pass or fail)
use a chi square analysis. If there are only 2 categories, you can also perform a phi-coefficient.
T/F: the sign of the phi-coefficient for 2 non-numerical variables is meaningful
false. only the strength is meaninful. you cannot determine if it is linear or the sign of the graph because the variables are just asigned numbers 0 or 1.
when would you use a phi-coefficient? a point biserial coefficient?
phi: when both variables of study in a correlational design are non numerical
point bi: when one variable of a correlational design is numerical, and the other isnt.
what is the coefficient of determination? example?
the r squared value.
measures how much variability in one variable is predictable in its relationship with the other varaible.
ex/ R squared value =0.64. 64% of the difference in GPA can be predicted by the differences in IQ.
in order for significance in the correlational value to be valid, the sample population size must be __
large.
requirement in order for the significance in the correlational value to be valid
need a large population size
in terms of prediction, correlational studies often identify one variable as the ____ and the second variable as the ____. give an example.
in terms of prediction, correlational studies often identify one variable as the PREDICTOR VARIABLE and the second variable as the CRITERION VARIABLE. give an example.
ex/ the correlational relationship between college performance and SAT score can predict how well a student will do in college (criterion variable) based on their SAT score (predictor variable)
What statistical process is used to allow correlations to be used for prediction? how is this laid out?
regression is used. Usually the y axis is the criterion, and the x axis is the predictor value.
example of an application of a correlational study that is not prediction
to learn about the unknown variable by seeing how it is related to an established variable.
ex/ looking at how genetic factors (known) interact with risk of alzheimers disease (complex and unknown)
how is reliability tested in correlational studies?
using the test and retest method
how is validity tested in correlational studies?
determined using concurrent validity. do the results of the new correlational test match up with expected results from an old, well established test?
a strong correlation should have high concurrent validity
a strong correlation should have ___ ____ validity
a strong correlation should have high concurrent validity
Applications of correlational studies?
1) making predictions via statistical regression
2) learning about the unknown
3) evaluating theories.
- theories can generate questions that can be answered via correlational designs. ex/ nature vs nurture. can use a correlational study with twins that were adopted to different parents. you cannot do manipulation though.
advantages of correlational design
can identify variables that might suggest further investigation
- high external validity
- can investigate variables that would be unethical to manipulate
disadvantages to correlational research design
- low internal validity. no controls
- third variable problem: there is a possibility that a 3rd unidentified variable is controlling the 2 other variables, and is responsible for producing the observed relationship.
- directionality problem. a correlational study does not determine which variable is the cause and which is the effect. it just points out that there is a relationship between two variables.
how do you statistically evaluate the relationship of a correlational design that uses more than 2 variables?
ANOVA, and multiple regression statistics to make predictions
ex/ studying IQ, motivation and academic performance. allows researchers to examine the relationship between specific variables while controlling the influence of other confounding variables.
- adding an additional predictor variable into the regression analysis adds to the prediction AFTER the influence of earlier predictors had already been considered.
what statistical tests can be done on a correlational study that has more than two groups?
ANOVA
what statistical tests can be done on a correlational study with 2 groups?
independent measures T test
