Research Methods: Lecture 5
|Correlational Research definition||The independent variable is not manipulated by the experimenter (just measured). Research relies on finding natural variation in the variables.|
|Important difference between correlational research and experimental research|| In correlational research, the independent variables normally correlate with each other. |
In experimental research the independent variables do not with each other or with background variables.
(Within-subjects, same backgrounds mean they can't correlate and between-subjects, random assignment means doesn't correlate)
|Positives with correlational research|| - Good EXTERNAL VALIDITY (as normally studies human behaviour in its natural environment)|
- Cheap and easy to do
|Negatives with correlational research||- Major problems with INTERNAL VALIDITY as cannot infer causation. (Variables often correlate with each other, something that experimental research gets around).|
|Collecting data for correlational research|| - Questionnaires|
- Official stats
- Surveys (psychological scales)
|Correlation coefficients|| - r (Pearson correlation coefficient)|
- p (Spearman rank)
Range from -1 to 1 with 0 being no correlation. Perfect relationships imply that no other variable is important for predicting an effect.
|Variance accounted for||Given by r^2 and can range from 0 to 1. (When r=1 we have accounted for all of the variance)|
|Simple linear regression|| (Least squares). Involves finding the lines that expresses the relationship between the two variables whilst producing the smallest squared error.|
Y'= A + BX
where Y is the dependent variable and X is the independent variable
|Standardized linear regression|| The dependent (Y) and independent (X) variables in the linear regression line can be standardized to give y= (beta)X. This means that the variables have had their MEANS SUBTRACTED from them and have been DIVIDED BY THEIR STANDARD DEVIATION. |
Beta= r for one variable.
|Idea of multiple regression|| A relatively simple linear regression equation with one independent variable turns into a more complicated equation with more than one independent variable (needed for everyday-life complexities).|
Y= A + B1X1 + B2X2... BnXn
y= (beta)1X1 + (beta)nXn
Describe the Y variable as regressed on the independent variable (X)
|Two reasons for doing multiple regression|| 1.) Generally get better prediction if you use more independent variables.|
2.) Increases internal validity as deals with the 3rd variable problem and therefore can talk of causation. It is therefore controlling for extraneous variables statistically (instead of random assignment).
|Beta weights are...||Measures of what the correlations would be if all the other variables in the equation were held constant. E.g. estimate of result if everone had same education level.|