- Technical Information
- Purpose of the tables
- References
CRBIND. Technical information
CRBIND. Purpose
The tables provides the unconditional critical region for
testing the independence in a 2´ 2 table (multinomial
case).
If the cell probabilities and the observed values for the four cells in
a 2´ 2 table are, respectively,
the aim is to test:
Ho º
OR =1 (independence or not association)
vs Ha º OR < 1 or Haº OR > 1 (one
tail)
or vs Ha º OR ¹
1 (two tails)
with OR= p11 p22 / p12 p21
the poblational odds-ratio.
Under Ho:
P(
n11, n12, n21, n22 | n..
) = n..! p.1n.1
p.2n.2 p1.n1.
p2.n2. /( n11!
n22! n12! n21!)
For a target error a the critical region is
a set, CR, of triplet (n11, n12, n21),
with n11+ n12+ n21 £
n.., so the error a is:
a (p1., p.1) = SCR
P( n11, n12, n21, n22
| n.. )
and de size of the test will be:
a *= Max a (p1.,
p.1) in 0< p1., p.1
< 1
This tables gives the CR for the Barnard's
test following the optimal procedure. The one-sided Ha considered is that
indicate by the data:
Ha º OR<1
if (n11 n22 / n12 n21)
< 1
---o0o---
CRBIND. References
For further details see:
- MARTÍN ANDRÉS, A. and TAPIA GARCIA, J. M. (1998). 'On
determining the P-value in 2x2 multinomial trials'. Journal of Statistical
Planning and Inferences 69(1), 33-49.
- MARTÍN ANDRÉS, A. and TAPIA GARCIA, J. M. (1999). 'Optimal
unconditional test in 2x2 multinomial trials'. Comput. Statis. &
Data Anal. 31(3), 311-321.
- TAPIA GARCIA, J. M. and MARTÍN ANDRÉS, A. (2000). ''Optimal
unconditional critical regions for 2x2 multinomial trials'. Journal
of Applied Statistics 27(6), 689-695..
|