|
SIMPLE AND COMPLEX
RANDOM ASSIGNMENT IN EXPERIMENTAL DESIGNS |
|
Valentim
Rodrigues Alferes, |
|
|
|
|
|
|
Example 1. The researcher has 3 treatments
and wants to perform a complete counterbalancing of treatments
order. He/she needs a minimum of 6 units or other multiples of 6 (12, 18, 24,
30, etc.; see General Recommendations). Only
23 units are available. In that case, we suggest that he/she runs CC3.sps, with 4 units per sequence, excluding from the
design the 24th sequence by inserting an additional syntax line: |
|
|
|
|
|
Example 2. The researcher has 5 treatments
and wants to perform a double carryover design. He/she needs a
minimum of 10 units or other multiples of 10 (20, 30, 40, etc.; see General Recommendations). Only 8 units are
available. In that case, we suggest that he/she runs CO5.sps,
excluding from the design the last two sequences by inserting an additional
syntax line: |
|
|
[...]
|
|
|
|
|
|
|
|
An alternative to truncated
counterbalancing is random ordering of
treatments made independently for each unit
included in the study. We don’t recommend this procedure (see General Recommendations). However,
random ordering of treatments can be done with IndRO.sps. |
|
|
|
|