Materials and Methods
The study population comprised all (479) dairy farms that were enrolled on Official (DHIA) testing programs in nine counties in Ohio. Five counties (Ashland, Holmes, Stark, Tuscarawas and Wayne) were in Northeastern Ohio, and four (Auglaize, Darke, Mercer and Shelby) in western Ohio.
A questionnaire designed to determine herd characteristics and management practices of the study herds was administered to the producers by DHIA supervisors in December, 1996 - January, 1997. Supplemental data regarding dairy herd characteristics pertaining to the eligible farms in the study area for the month of January 1997 were obtained from DHIA and utilized in the analysis.
Data from returned surveys and from Ohio Dairy Herd Improvement Cooperative, Inc. (DHIA, Powell, OH) records were entered into a spreadsheet (ExcelÒ, Microsoft Corp., Redmond WA), and later transferred to a SAS/STAT (version 6.12, SAS Institute, Inc., Cary, NC) program for further statistical analysis. Descriptive statistics (means, medians, standard deviations and ranges) were calculated for continuous variables and frequency distributions evaluated for categorical variables.
Regarding association of herd characteristics/management practices with milk production, data were analyzed using a multivariable ANOVA procedure. All independent variables (herd characteristics/management practices) were first screened for their simple association with the dependent variable (milk production as measured by RHA) based on the F-statistic. Variables that met a critical p-value of 0.25 on the initial screening were used for further multivariable modelling. Multivariable model building employed a forward selection stepwise procedure at P = 0.05 similar to that described in the literature (Hosmer and Lemeshow, 1989).