proc plan incomplete block design

[vstorme]

Hej,

I try to design an incomplete block design for 13 replicates of 213 different plant varieties (considered as treatments) with block size 71. I tried following code:

title 'Generalized Cyclic Block Design';
proc plan seed=33373;
   treatments Treatment=71 of 213 cyclic (  3	 7   9	 10	 14	 19	 23	
                                           30	33	39	 43	 54	 56	 58	
                                           62	63	64	 71	 73	 75	 83	
                                           88	89	90	 95	 99	100	101	
                                          104  109 110	112	113	114	117	
                                          119  121 122	124	126	129	133	
                                          136  139 142	147	148	151	152	
                                          153  157 160	161	162	168	173	
                                          174  176 179	180	185	187	188	
                                          189  193 200	204	206	210	211	213) 3;
   factors Block=39 Plot=71;
   output out=GCBD;
quit;

I do not get the same replicate number for each variety (treatment). I also tried:

title 'Generalized Cyclic Block Design';
proc plan seed=33373;
   treatments Treatment=71 of 213 random;
   factors Block=39 Plot=71;
   output out=GCBD;
quit;

The SAS example gives a perfect replicate number of 2:

title 'Generalized Cyclic Block Design';
proc plan seed=33373;
   treatments Treatment=8 of 52 cyclic (1 2 3 4 32 43 46 49) 4;
   factors Block=13 Plot=8;
   output out=GCBD;
quit;

Does it have to do with how you create the initial block and the step?

with kind regards,

Veronique

[data_null__]

How did you come up with the initial block?

[vstorme]

I just sampled 71 values without replacement from 1:213 and put them in increasing order

[StatDave]

See the examples of constructing incomplete block designs in the PROC OPTEX chapter of the SAS/QC User's Guide.

[vstorme · Accepted solution]

Thanks Dave. Based on the paper from Hans-Peter Piepho I tried the following successfully:

title 'resolvable IBD' ;
* example: v=30 treatments, block size k=5, r=3 replicates;
* nr of blocks needed per rep = 30/5 =6 ;
%let v=30;
%let k=5;
%let r=3;
%let b=6;

data TreatmentLabels;
  do trt=1 to &v;
    output;
  end;
run;

data layout;
  do rep=1 to &r;
    do block=1 to &b;
      do plot=1 to &k;
	    output;
      end;
    end;
  end;
run;

proc optex data=TreatmentLabels seed=98221534;
  class trt;
  model trt;
  blocks design=layout niter=10000 keep=10;
  class rep block plot;
  model rep, block(rep) / prior= 0, 10;
  output out=RIBD;
run;
quit;

This however takes a very long time, and I need v=213, k=71, r=13, b=3. It still was not finished after 12 hours.