Bar charts are often a better approach especially if several of the values displayed are similar in magnitude as the human eye can actually detect pretty small variations of the tops of the bars.
To make multiple charts, of which every style you choose you could use SGPLOT with BY group processing for which requires the data to be sorted by those variables. Or SGPANEL with Panelby though that takes a bit more work but places all the graphs in a single image is specified correctly.
The vbar (or hbar) in sgplot lets you specify a category variable and optionally a response variable (such as expenditure), a statistic such as Freq, Mean, Median, Percent or SUM which will be applied to the response variable (mean, median and sum kind of require a response variable) and a group variable to display the sum for sub-groups of your category variable. The groupdisplay option controls how the subgroups are used on the graph, stack places them one above the other and cluster side-by-side.
so you could do something like this in SGPLOT
vbar quarter /response=value stat=sum group=expenditure groupdisplay=stack would have one bar per quarter with the values summed by expenditure type.
If you do a BY trichot you would get one bar chart be level of the trichot variable.
Or as I mentioned SGPANEL.
It is also simple to create a second graph by changing role of expenditure and quarter so that for the same type of expenditure you would a sum per quarter which may actually be easier to interpret as one might expect the relative quarter-to-quarter expenditures to be similar.
... View more
