Skip to main content
ARS Home » Research » Publications at this Location » Publication #76875
Title: EMPIRICAL ESTIMATES OF POWER OF THE TEST FOR DETECTING TREATMENT DIFFERENCES WITH BINOMIAL DATA

Author
item SPLAN, R. - UNIV. OF NEBRASKA-LINCOLN
item HAFS, H. - RUTGERS UNIVERSITY
item Van Vleck, Lloyd

Submitted to: American Society of Animal Science
Publication Type: Abstract Only
Publication Acceptance Date: 2/21/1997
Publication Date: N/A
Citation: N/A

Interpretive Summary:

Technical Abstract: The objective was to develop a method to determine numbers of locations and animals per treatment per location to achieve a specified power of test for an experiment measuring return to estrus in anestrus postpartum cows. Estimates by REML of among location (.012) and total variance (.178) were obtained from a similar data set and then used to generate simulated data for the binomial trait. Pre-determined factors were locations (4, 6, 10 or 20), animals per treatment per location (25, 50, 75 or 100), difference due to treatment (plus .00, .05, .10, .15, .20 or .25 from a control of .50), alpha probability level (.010, .025, .050 or .100) and ratio of among location to total variance (.05, .10 or .20). Each combination was replicated 1000 times and number of significant treatment differences was counted. Fraction of replicates with significant differences is an empirical estimate of power of test. Due to number of combinations, only some results are given. With ratio of among location variance to total variance of .05, six locations, .05 alpha level and .10 difference due to treatment, power was 65.4, 87.7, 94.7 and 98.0% for 25, 50, 75, and 100 animals per treatment per location. With ratio of among location to total variance of .05, 50 animals per treatment per location, .05 alpha level and .10 difference due to treatment, power was 72.3, 87.7, 96.3 and 100% for 4, 6, 10 and 20 locations. With ratio of among location to total variance of .05, six locations, 50 animals per treatment per location and .05 alpha level, power was 46.5, 87.7, 99.3 and 100% for differences due to treatment of .05, .10, .15, and .20. Results indicate this method can determine power empirically and animals needed in experiments with binomial measurements.