introduction to spss RESEARCH METHODS & STATISTICS HANDBOOK PHẦN 5 docx

41

the mean age is 31.77). The next table contains the correlation (Pearson‟s) for the two

variables, just as if you had run the correlation procedure. The coefficient is -0.57, so

it is fairly high and is negative (as one goes up, the other decreases).

For the meantime, ignore the table labelled „variables entered/removed‟.

The next important table is “Model Summary”. The R and R-squared values are

given for the equation (0.571, as above, and 0.325). Don‟t worry too much about the

other values in this table.

The next table contains the regression ANOVA. This test indicates how good the

model is - whether there is some overall relationship between the dependent and

independent variable(s). The key element is the F score. For this regression, the F

score has an associated p value of 0.017, well below the .05 cut-off. This indicates

that there is a linear relationship. It should be noted however that only an

examination of the scatter plot of the variables can confirm that the relationship

between two variables is linear.

The next table contains some really important information! The table is labelled

“Coefficients” and contains the regression equation. The regression coefficient and

constant are given in column B of the table. The equation therefore is:

Predicted height to weight ratio = -.00368(AGE) + .602

The t value indicates whether each independent variable has a significant individual

impact on the regression equation. In simple regression, there is only one independent

variable, and, for this one, it has a significant influence (a t score with an associated p

value of 0.0168 - notice it‟s the same as the ANOVA score).

The next section begins with “Residual Statistics.” This gives means, SDs and other

information about the unstandardised and standardised predictor and residual scores in

the regression.

You could follow up the regression by doing up a scatter plot. Look at your scatter

plot. Basically, all you need to know is that if the plot shows no obvious pattern than

this confirms that the assumptions of linearity and homogeneity of variance have been

met. Where you get into trouble is if the points form a crescent or funnel shape. If

this is the case, further screening of your data is necessary.

Multiple Regression

Often, it is too simplistic to assume that a single independent variable is all that is

required to make some sort of prediction about the scores for a dependent variable.

This is where you have to run multiple regression.

For now, the regression will look at the impact of age (AGE), height to weight ratio

post-plan (HWRATIO2) and height to weight ratio long after the plan (HWRATIO3)

on the dependent variable, the subjects‟ initial height to weight ratio (HWRATIO). To

run the analysis, choose: ANALYSE, REGRESSION and then LINEAR.