topic Re: Experimental Design GRBD in SAS Procedures

Experimental Design GRBD

deleted_user — Thu, 03 Sep 2009 16:30:06 GMT

Hi SAS Procedures Group,

I am needing some help in analyzing an experiment in a vineyard in a generalized randon block design. I have 3 blocks, 6 light treatments and I have separate results from both sides of the grapevines (east and west). Each light treatment is replicated two times within each block, so I have a total of 36 plots. I have as response variable the sugar content of the fruit. I would like to know if someone could help me with the sas statements to analyze these experiment.
In advance thank you very much

Re: Experimental Design GRBD

data_null__ — Thu, 03 Sep 2009 17:21:44 GMT

This is the question that you should have the answer to before you apply any treatments, collect any data or spend any money. 🙂

I'm not a statistician but I know when I need one.

I'm not sure the following is correct, but it seems reasonable. Perhaps a statistician will offer some advice.

[pre]
/* Make sample data */
proc plan seed=1150759636;
factors
block = 3 ordered
light = 6 random
side = 2 ordered
plot = 1 of 36 perm
;
treatments sugar = 1 of 500;
output out=plan side cvals=('E' 'W');
run;
proc print;
run;
proc glm;
class block light side;
model sugar = block light block*light side(block);
test h=light e=block*light;
run;

proc mixed;
class block light side;
model sugar = light;
random block block*light side(block);
lsmeans light / pdiff;
run;
[/pre]

Re: Experimental Design GRBD

deleted_user — Thu, 03 Sep 2009 18:00:52 GMT

Dear Data_null,
Thank you very much. I am missing in your porgramathe replication within the block. I have really 72 data, because I have 6 treatments, 3 blocks, 2 replicate of the same light treatment, and both sides of the canopy, East and West. The other question I have, but maybe as you said I have to ask a statistician, is how worry I have to be here about normality and homegeinity of variance.
Maybe you have a program in order to consider the 72 data.

Thanks again

Re: Experimental Design GRBD

data_null__ — Thu, 03 Sep 2009 18:08:51 GMT

> 2 replicate of the same light treatment,

You say you have 36 plots. Does this imply 72 plots? Or 2 samples on the sample of the same plot.

You will need to "draw a picture" I don't understand

Re: Experimental Design GRBD

data_null__ — Sat, 05 Sep 2009 01:43:26 GMT

This seems reasonable but you really need to get a statistician to help you...

[pre]
proc plan seed=1150759636;
factors
block = 3 ordered
rep = 2 ordered
light = 6 random
plot = 1 of 36 perm
side = 2 ordered
i = 1 of 72 perm
/ noprint
;
treatments sugar = 1 of 500;
output out=plan side cvals=('E' 'W');
run;
proc print;
run;
proc glm;
class block rep light side;
model sugar = block light block*light block*light*rep side side*block;
test h=light block*light e=block*light*rep;
run;

proc mixed method=type3;
class block rep light side;
model sugar = light block*light side;
random block block*light*rep side*block;
lsmeans light / pdiff;
run;
[/pre]