Longitudinal data are data containing measurements on subjects at multiple
times. Visualizing longitudinal data without loss of data can be difficult, but
it is possible to do so in SAS. Once your dataset is in the appropriate
configuration, proc gplot allows you to generate plots with time on the
horizontal axis and levels of an outcome on the vertical axis. Graphing options
allow you connect the
related points (points from a single subject or from a specified group) with a
line. This type of plot allows you to trace the levels of the outcome variable
over time for a given subject and can often reveal larger patterns that may be
of interest.
If your dataset is "wide" (each time point is represented as its own variable
and the measured responses are the values found in the multiple time variables),
then you will need to convert your data to long form in order to visualize your data
using proc gplot.
Obs id time1 time2 time3 time4 1 1 23 25 25 30 2 2 14 17 20 23 3 3 32 33 34 24 4 4 33 35 42 50 5 5 5 17 22 22
The dataset above contains four measurements on five subjects in wide form. For details on how to reconfigure such a
dataset to long form, see the "Reshaping Wide to Long" section of the SAS
arrays seminar. Below, we have the same dataset in long form.
Obs id time y 1 1 1 23 2 1 2 25 3 1 3 25 4 1 4 30 5 2 1 14 6 2 2 17 7 2 3 20 8 2 4 23 9 3 1 32 10 3 2 33 11 3 3 34 12 3 4 24 13 4 1 33 14 4 2 35 15 4 3 42 16 4 4 50 17 5 1 5 18 5 2 17 19 5 3 22 20 5 4 22
If you dataset is "long" (there is one variable indicating time and another
variable with the measured response from the given time value), then proc gplot
will allow you to visualize your data without reconfiguring it. We can indicate
the variables represented on our vertical and horizontal axes as we would in an ordinary gplot, and
then indicate how the records are grouped with = (groupvar). In this
case, we are interested in grouping the records by id. In order to join the
points associated with one id, we need to indicate "join" as part
of our symbol options.
symbol1 value = circle color = black interpol = join repeat = 5; proc gplot; plot y*time = id / nolegend; run;
We can see the plot
generated by the above code includes all of the information in our dataset with
one line for each id. For
more ways to plot longitudinal data in SAS, see Chapter 2 of Modeling
Longitudinal Data and Chapter 2 of Applied Longitudinal
Data Analysis.