Detail On the Choice of Design and of Test Statistic in the Analysis of Experiments with Sampled Materials   Oleh:  Wickens, Thomas D. ; Keppel, Geoffrey Jenis: Article from Bulletin/Magazine Dalam koleksi: Journal of Verbal Learning and Verbal Behavior vol. 22 no. 3 (Jun. 1983), page 296-309. Ketersediaan Perpustakaan PKBB Nomor Panggil: 405/JVL/22 Non-tandon: tidak ada Tandon: 1  Lihat Detail Induk Isi artikel The failure to treat samples of materials as random effects in the analysis of variance suggests that type I errors abound. However, a proficient experimenter probably handles materials more subtly than a superficial model implies. A series of simulated experiments is used to assess the consequences of selecting material samples in various ways. When material variability is present and samples are selected haphazardly, tests based on the quasi-F statistics are correct, although any such design lacks power unless sufficiently large samples of material are used. If materials are sampled systematically, many of the diffculties are reduced. Where the size of the material variation is not too great and can be accurately balanced across conditions, quasi-F tests that include this blocking are most powerful and least error prone. However, if the blocking is ignored, quasi-F tests have substantial negative bias, while the conventional fixed-effect F, statistic can give more powerful tests while incurring only limited vulnerability to false-positive errors. Opini Anda Klik untuk menuliskan opini Anda tentang koleksi ini! Kembali

Copyright © 2006, 2007 Unika Atma Jaya, all rights reserved