* - FILE: Humfacs.sas;
* repeated measures example--human factors research. Subjects
were given a computer aided instructional program at one sitting.
The program presented text and then asked a series of questions
about the text, providing appropriate feedback and review for
incorrect responses to the test questions. There were five
such instructional sets. The dependent variables are the number
of error in each instructional set. They are called INSTR1 through
INSTR5 according to the order of presentation. From previous
research, the level of difficulty had been calibrated to be
the same for each instructional set. Two groups of subjects
were used. The first completed the instructional program on
a standard black/white computer display. The second completed
the program on a color display with borders, highlighted text,
etc. The purpose of the study was to determine whether it was
worth the time and effort to write programs with all the
"bells and whistles.";
* PS. I haven't the vaguest idea if these data are in any way
realistic. I just made them up.;
OPTIONS NOCENTER NODATE PAGENO=1 PAGESIZE=60 LINESIZE=76;
DATA HUMFACS;
LENGTH GROUP INSTR1-INSTR5 3 DEFAULT=4;
INPUT GROUP INSTR1-INSTR5;
LENGTH TYPE $8;
TYPE='B & W'; IF GROUP=2 THEN TYPE='COLOR';
DATALINES;
1 11 18 17 21 18
1 8 10 16 17 16
1 8 15 14 11 17
1 13 15 17 18 18
1 13 19 18 19 19
1 17 14 23 19 20
1 17 17 22 17 15
1 12 16 15 15 15
1 13 14 15 14 16
1 17 19 16 21 18
2 10 11 13 11 16
2 21 22 20 22 22
2 9 11 11 10 13
2 16 16 15 15 17
2 11 12 15 15 17
2 13 13 14 13 16
2 11 9 11 12 12
2 11 10 15 14 16
2 16 12 14 20 16
2 13 10 16 21 23
RUN;
TITLE 'PSYCHOLOGY 7291: REPEATED MEASURES ANOVA: HUMAN FACTORS EXAMPLE';
PROC PRINT; BY TYPE;
VAR INSTR1-INSTR5;
RUN;
PROC MEANS; BY TYPE;
VAR INSTR1-INSTR5;
run;
* this GLM performs a repeated measures ANOVA and a repeated
measures MANOVA. A HELMERT transformation of the dependent
variables is requested. This compares the number of errors
in an instructional set with the average number of errors in
all the subsequent instructional sets. This is a useful
transformation when: (1) you know the means should all be
the same if the instructional sets were presented in
random order, and (2) you want to examine inprovement over
time or decreased performance over time.;
PROC GLM;
CLASS type;
MODEL INSTR1-INSTR5=type;
TITLE2 Traditional MANOVA;
MANOVA H=type;
RUN;
* - options PRINTM and PRINTE print, resepctively, the
model and the error matrices. SUMMARY requests a
univariate ANOVA for each of the transformed
variables;
TITLE2 Repeated Measures ANOVA;
REPEATED INSTRSET 5 HELMERT /PRINTM PRINTE SUMMARY;
RUN;