SAS, unbalanced 2-way MANOVA. Chip manufacture.

Examine whether pressure and power are two main explanatory variables to uniformity (UNIF1 and UNIF2) and selectivity (SELECT), where SELECT = etch1/ etch2.

data etch;

input press power etch1 etch2 unif1 unif2;

selecti = etch1/etch2;

power_sq = power**2;

datalines;

240 90 793 300 13.2 25.1

240 90 830 372 15.1 24.6

240 90 843 389 14.2 25.7

240 110 1075 400 15.8 25.9

240 110 1102 410 14.9 25.1

240 130 1060 397 15.3 24.9

240 130 1049 427 14.7 23.8

290 90 973 350 7.4 18.3

290 90 998 373 8.3 17.7

290 110 940 365 8.0 16.9

290 110 935 365 7.1 17.2

290 110 953 342 8.9 17.4

290 110 928 340 7.3 16.6

290 130 1020 402 8.6 16.3

290 130 1034 409 7.5 15.5

;

run;

proc glm data = etch;

class press power;

model selecti unif1 unif2 = press power press*power/ ss1 nouni;

manova h = press power press*power/ printe printh;

run;

Type1 analysis for model building. "press" precedes "power" by presuming "press" is more influential.

No interaction effect.

Significant Press effect. Weak Power effect.

title1 j=left "test for quadratic relationship: power_sq";

proc glm data = etch;

model selecti unif1 unif2 = press power power_sq/ss1 nouni;

manova h = press power power_sq/ printe printh;

run;

Insignificant quadratic relationship of power to response.