1

Putting a regression line on a scatterplot

Putting a regression line on a scatterplot in SPSS can be done using the interactive graph in SPSS. The syntax below draws a regression line using reading score to predict writing score. Fuller details are given here.

IGRAPH
 /X1 = VAR(READ)
 /Y = VAR(WRITE)
 /FITLINE METHOD = REGRESSION LINEAR LINE = TOTAL
 /SCATTER.

Alternatively you can run the following syntax which produces a scatterplot:

GRAPH
  /SCATTERPLOT(BIVAR)=READ WITH WRITE
  /MISSING=LISTWISE .

To add the fit line, you need to open the graph editor (by double-clicking on the graph), and then select the Elements pull-down menu at the top, and then select Fit Line at Total.

You just need to change the variable names in the above syntax examples substituting the variable names in your SPSS data for 'read' and 'write'.

If you wish to produce separate regression lines for individual subgroups defined by a factor e.g. group then the above syntax is easily adjusted as the below

GRAPH
  /SCATTERPLOT(BIVAR)=READ WITH WRITE BY GROUP
  /MISSING=LISTWISE .

On the Elements pull-down menu (referred to as above) select Fit Line at Subgroups.

Suppose you wish to add user defined lines to a plot e.g. to represent the ANCOVA assumption of equal slopes within various groups. You can do this by running the syntax below which re-expresses the earlier syntax for variables in a different data set.

GRAPH
  /SCATTERPLOT(BIVAR)=FAC1_4 WITH WCSTtotalscore by GroupH_L
  /MISSING=LISTWISE .

You can make adjustments using the graph editor (at least in version 13 of SPSS, probably also in later ones) by double clicking on the graph then choosing options>reference line from equation and inserting the equation for the line you wish to add. If you then click on the 'lines' tab you can edit the colour and form of the line. Click apply and close. Repeat for any other lines you wish to add which will be appended to the plot.

You should end up with something like the plot on slide two of this powerpoint file which added three lines (one for each group defined by the GroupH_L factor) to represent the relationship in each group between FAC1_4 (the covariate) and WCSTtotalscore (the outcome). The equations of the lines are obtained using the regression parameter estimates (presented on the first slide of this powerpoint file with the plot) which were produced from running the ANCOVA procedure using the syntax below:

UNIANOVA
  WCSTtotalscore  BY GroupH_L  WITH FAC1_4
  /METHOD = SSTYPE(3)
  /INTERCEPT = INCLUDE
  /PRINT = PARAMETER
  /CRITERIA = ALPHA(.05)
  /DESIGN = FAC1_4 GroupH_L .

The ANCOVA table is given on the third slide in the powerpoint file.