Help with SAS Anova codes

[jmoseley]

I am trying to figure out what I am doing wrong with this ANOVA code. Help please.

question

A researcher was interested in the influence of test administration practices (in-class testing vs. take-home testing) on student anxiety. A volunteer sample of 18 undergraduate psychology students participated. The students were pretested with a trait anxiety instrument, and on the basis of their scores, were classified as High Anxious or Low Anxious. Students within each anxiety category were then randomly assigned to receive either a take-home midterm examination or an in-class midterm examination. The researcher hypothesized that the ‘impact’ of test type would be greatest for the high anxiety students. On the day of the examination, the researcher administered a test anxiety instrument to obtain a measure of the students’ anxiety about taking the midterm. The following test anxiety scores were obtained:

Student Trait Anxiety	Type of Test
In-Class	Take-home
Low	22 16 21 17	20 18 22
High	21 35 33 25 25 28	24 23 19 21 22

data case3;

input trait $ 1 test 3 score 5-7;

datalines;

L 22

L 16

L 21

L 17

H 21

H 35

H 33

H 25

H 25

H 28

L 20

L 18

L 22

H 24

H 23

H 19

H 21

H 22

;

PROC ANOVA;

CLASS TRAIT TEST;

MODEL SCORE = TRAIT TEST TRAIT*TEST;

MEANS TRAIT TEST/TUKEY;

RUN;

[Rick_SAS]

After the DATA step, put

PROC PRINT; RUN;

You will see that although your description and the PROC ANOVA code use three variables, you are only creating two variables in the DATA step.

[jmoseley]

Thank you for your insights Rick. I tried to change the data step to trait $ test $ class home; but it is giving me tons of other errors. By the way when I input the full question on the help board, it somehow failed to enter in-class and take-home data in two separate columns.

[Rick_SAS]

I think you need to code the in-class and take-home variable.

Maybe use 'C' for in-class and 'T' for take-home.

[Rick_SAS]

Maybe like this?

data case3;
input trait $1 test $3 score;
datalines;
L C 22
L C 16
L C 21
L C 17
H C 21
H C 35
H C 33
H C 25
H C 25
H C 28
L T 20
L T 18
L T 22
H T 24
H T 23
H T 19
H T 21
H T 22
;

I notice that this is an unbalanced design (some groups are different sizes than others).  For these data you'll want to use PROC GLM instead of ANOVA.

[jmoseley]

Thank you very much Rick, I will try this...Josie

[jmoseley · Accepted solution]

I tried those codes Rick, and all it is giving me is a frequency table. These are my codes.
data case3;
input trait $ 1 test $ 3 count;
datalines;
L C 22
L C 16
L C 21
L C 17
H C 21
H C 35
H C 33
H C 25
H C 25
H C 28
H T 20
H T 18
H T 22
H T 24
H T 23
H T 19
H T 21
H T 22
;
PROC GLM data=case3;
CLASS TRAIT TEST;
MODEL SCORE = TRAIT TEST;
RUN;
proc print;
run;

[Ksharp]

data case3;
input trait $  test $  count;
datalines;
L C 22
L C 16
L C 21
L C 17
H C 21
H C 35
H C 33
H C 25
H C 25
H C 28
L T 20
L T 18
L T 22
H T 24
H T 23
H T 19
H T 21
H T 22
;
run;
/*Check mean std for each cell*/
proc tabulate data=case3;
class trait test;
var count;
table trait*test,count*(mean std n);
run;
/*Check main effect*/
proc glm data=case3;
class trait test;
model count=trait test/ss1 ss3;
run;
proc glm data=case3;
class trait test;
model count=test trait /ss1 ;
run;
/*Check full model*/
proc glm data=case3;
class trait test;
model count=trait|test/ss1 ss3;
run;

Source	DF	Type I SS	Mean Square	F Value	Pr > F
trait	1	137.1544012	137.1544012	10.13	0.0067
test	1	48.9186983	48.9186983	3.61	0.0782
trait*test	1	52.0713450	52.0713450	3.84	0.0701

Source	DF	Type III SS	Mean Square	F Value	Pr > F
trait	1	119.0187135	119.0187135	8.79	0.0103
test	1	26.6678363	26.6678363	1.97	0.1824
trait*test	1	52.0713450	52.0713450	3.84	0.0701

Both kind of ANOVA(SS1 SS3) show trait is significant , test and trait*test - (interactive effect) are NOT significant.

So there is NO significant influence of test administration practices (in-class testing vs. take-home testing) on student anxiety.

[jmoseley]

Thank you again very much for your assistance Mr. Keshan, it is very much appreciated.