请教：关于一个程序的结果问题

shiyiming

某研究所研制了三个降血脂中药复方制剂，现拟对三个复方与标准降脂药（安妥明）的疗效进行比较。取品种相同，健康的雄性家兔16只，按其体重大小分四个配伍组，每个配伍组中的四只家兔被完全随机地分入四个药物组中接受实验。各药物组的动物均饲以同样高脂饮食，并每日分别灌以不同药物，第45天处死动物，观察其冠状动脉根部动脉粥样硬化斑块大小。研究数据如下
比较不同药物对动脉粥样硬化斑块形成的面积大小的影响之间是否有差别？


data aa;
do block=1 to 4;
do drug=1 to 4;
input x@@;
output;
end;
end;
cards;
0.000 0.283 0.114 0.094
0.009 0.196 1.146 0.131
0.003 0.217 0.158 0.065
0.001 0.236 0.159 0.087
;
run;
proc anova data=aa;
class block drug;
model x=block drug;
means drug/dunnett('1');
run;
如果我把上面的means语句后面的选项改为snk，即means drug/snk;然后结果就大相径庭，大侠看看是怎么回事啊？为什么会这样？

shiyiming

I don't see any difference in two SAS outputs.

Two F-tests gave the same p-value.

The output for the multiple comparisons may looks different due the two different options you used in the program, but they gave the same conclusion. Dunnett compare drug 2,3 4, with drug 1, while SNK compare each possible pairs among four drugs.

Please read the ANOVA chapter in SAS help for more information.

Hope this helps.