kingang
有三个不同的处理组,三个处理组又分三个不同的剂量水平{高中低},高中低这三个剂量水平有阳性 阴性的结果,要求这三个处理组间是否有显著性差异?结果如下:
组别 水平 阳性 阴性
6 1 53 4947
6 2 69 4931
6 3 87 4913
3 1 46 3954
3 2 67 4933
3 3 79 4921
7 1 47 2953
7 2 27 1973
7 3 32 3968
neige
your GROUP can be fixed or random,
LEVEL can be fixed, random or nested within group
so you have to be more specific....
and you want to detect diff for 阳性 or 阴性 or together as a whole?
kingang
The groups are random , there are not control group ;
I want to detect diff for the three groups ;
I don't know that if i merge the results of neg and postive by group ,and then detect diff for
the three groups ,whether there exist information missing or not ;
I design the programm by SAS as follow :
data b36;
do r=1 to 2;
do c=1 to 2;
input x@@;
output;
end;
end;
cards;
209 14791
192 13808
;
proc freq;
weight x;
tables r*c/chisq norow nocol nopct;
run;
quit;
data b67;
do r=1 to 2;
do c=1 to 2;
input x@@;
output;
end;
end;
cards;
209 14791
106 8894
;
proc freq;
weight x;
tables r*c/chisq norow nocol nopct;
run;
quit;
data b37;
do r=1 to 2;
do c=1 to 2;
input x@@;
output;
end;
end;
cards;
192 13808
106 8894
;
proc freq;
weight x;
tables r*c/chisq norow nocol nopct;
run;
quit;
THANK YOU VERY MUCH! neige.
neige
what are you trying to do in ur program? read the data?
input group level pos neg @@;
datalines;
6 1 53 4947
6 2 69 4931
6 3 87 4913
3 1 46 3954
3 2 67 4933
3 3 79 4921
7 1 47 2953
7 2 27 1973
7 3 32 3968
kingang
Today ,I design another program,but i can get the national results .
THe program as follow;
data a;
input group level effect count@@;
cards;
6 1 1 53 6 1 0 4947
6 2 1 69 6 2 0 4931
6 3 1 87 6 3 0 4913
3 1 1 46 3 1 0 4954
3 2 1 67 3 2 0 4933
3 3 1 79 3 3 0 4921
7 1 1 47 7 1 0 4953
7 2 1 27 7 2 0 1973
7 3 1 32 7 3 0 3968
;
proc freq ;
weight count;
tables group*level*effect/chisq cmh norow nocol nopct;
proc print;
run;
quit;
THanks !
kingang
[quote]引用第3楼neige于2007-06-01 10:29发表的“”:
what are you trying to do in this program? read the data?[/quote]
3 6 7 这三个组间两两进行卡方检验!不过我把每个组的三个水平没考虑,阴阳性直接合并了,不知道这样做是否有信息损失。目的是,这三个间是否存在显著差异!
kingang
非常感谢!neige ,
neige
of course you can not do that!
if your group is random, then I do not think chisq is a good test, do you know anything on general linear model or mixed model?
neige
[quote]引用第7楼neige于2007-06-01 10:41发表的“”:
of course you can not do that!
if your group is random, then I do not think chisq is a good test, do you know anything on general linear model or mixed model?[/quote]
I am not sure what should you do since you have to responses and you do not know which one you are looking at or both, maybe you can try 2-way manova with one factor random. I am pretty sure GLM can handle multivariate, I do not think MIXED can do it, but did not try yet,
anyone?
kingang
[quote]引用第3楼neige于2007-06-01 10:29发表的“”:
what are you trying to do in ur program? read the data?
input group level pos neg @@;
datalines;
6 1 53 4947
.......[/quote]
neige ,您好,这样输入数据的话,我不知道下面该怎么编啦,还望不胜赐教。在统计方面我还是个新手,谢谢!
kingang
[quote]引用第8楼neige于2007-06-01 10:47发表的“”:
I am not sure what should you do since you have to responses and you do not know which one you are looking at or both, maybe you can try 2-way manova with one factor random. I am pretty sure GLM can handle multivariate, I do not think MIXED can do it, but did not try yet,
anyone?[/quote]
I will study the statistical methods , thank you for your good advice ;
but i don't know the anova method may use in the frequceny material!
neige
Do some plots to get some idea of the data, then decide what to do. your levels may be interacting with your groups, try this
goptions reset=all ftext=swissb htext=1;
axis1 width=2 offset=(3 pct) label=(a=90 r=0 font=swiss height=2.0 ) order=(20 to 100 by 10);
axis2 width=2 offset=(3 pct) label=(font=swiss height=2.0 ) order=(1 2 3);
legend1 across=1 cborder=black position=(bottom inside middle) offset=(0,10)
mode=share label=none value=(tick=1 '5') shape=symbol(6,2);
proc gplot data=test;
plot pos*level=group/vaxis=axis1 haxis=axis2 legend=legend1;
symbol1 v=dot i=join w=2 l=1 c=red ;
symbol2 v=circle i=join w=4 l=2 c=blue;
symbol3 v=circle i=join w=4 l=3 c=green;
run;
goptions reset=all ftext=swissb htext=1;
axis1 width=2 offset=(3 pct) label=(a=90 r=0 font=swiss height=2.0 ) order=(2000 to 5000 by 500);
axis2 width=2 offset=(3 pct) label=(font=swiss height=2.0 ) order=(1 2 3);
legend1 across=1 cborder=black position=(bottom inside middle) offset=(0,10)
mode=share label=none value=(tick=1 '5') shape=symbol(6,2);
proc gplot data=test;
plot neg*level=group/vaxis=axis1 haxis=axis2 legend=legend1;
symbol1 v=dot i=join w=2 l=1 c=red ;
symbol2 v=circle i=join w=4 l=2 c=blue;
symbol3 v=circle i=join w=4 l=3 c=green;
run;
losttemple
。。。。。。。。。
neige
不是我弄错了吧