THE RANDOM TOPOLOGY MODEL OF CHANNEL NETWORKS: BIAS IN STATISTICAL TESTS.

Direct tests of the random topology model have traditionally employed the 0.05 level of significance. As exercises in model confirmation, these tests should have used a larger significance level and because they did not, the test have been biased in favor of the model. This bias has contributed to t...

Descripción completa

Detalles Bibliográficos
Publicado en:	Professional Geographer Vol. 38; no. 1; pp. 77 - 82
Autores principales:	Abrahams, Athol D., Mark, David M.
Formato:	Artículo
Publicado:	Taylor & Francis Ltd Feb86
Materias:	Geographic network analysis Geography > Methodology Topology Relief models Geography > Statistical methods Statistical bias
Acceso en línea:	Ver este registro en EBSCOhost

Descripción
Sumario:	Direct tests of the random topology model have traditionally employed the 0.05 level of significance. As exercises in model confirmation, these tests should have used a larger significance level and because they did not, the test have been biased in favor of the model. This bias has contributed to the limitations of the model being overlooked. When all published tests of topologically distinct channel networks and ambilateral classes are repeated using a significance level of 0,30 rather than 0.05, the proportion of tests in which the model is rejected increases from 15 to 46 percent.