In certain areasof animal research, such as nutrition, quantitative summarizations of literaturedata are periodically needed. In such instances, statistical methods dealingwith the analysis of summary data (generally from the literature) must be used.These methods are known as meta-analyses. The implementation of a meta-analysisis done in several phases. The first phase defines the study objectives andidentifies the criteria for selecting prior publications to be used in the constructionof the database. Publications must be scrupulously evaluated for their qualitybefore being entered into the database. During this phase, it is important tocarefully encode each record with pertinent descriptive attributes (experiments,treatments, etc.) to serve as important reference points later on. Statistically,databases from literature data are inherently unbalanced, leading to considerableanalytical and interpretation difficulties. Missing data are frequent, and dataare not the outcomes of a classical experimental system. A graphical examinationof the data is useful in getting a global view of the system as well as to hypothesizespecific relationships to be investigated. This phase is followed by a statisticalstudy of the meta-system using the database previously assembled. The statisticalmodel used must follow the data structure. Variance decomposition must accountfor inter-and intra-study sources; dependent and independent variables mustbe identified either as discrete (qualitative) or continuous (quantitative).Effects must be defined as either fixed or random. Often observations must beweighed to account for differences in the precision of the reported means. Oncemodel parameters are estimated, extensive analyses of residual variations mustbe performed. The roles of the different treatments and studies in the resultsobtained must be identified. Often, this requires returning to an earlier stepin the process. Thus, meta-analyses have inherent heuristic qualities that canguide in the design of future experiments as well as aggregating prior knowledgeinto a quantitative prediction system.
