Data analysis and model validation in designed experiments and regression studies for agricultural experimentation
1998
Girma Taye (IAR, Addis Abeba (Ethiopia))
The paper discusses one problem area in the statistical analysis of data, namely assumptions underlying tests of hypothesis, and the need for statistical analysis of data in the context of Ethiopian Agricultural Research. Emphasis has been given to assumptions of additivity, homogeneity, normally and independence of error in both designed experiments and regression studies. Some mis-use and abuse of statistical methods by subject matter experimenters regarding model selection, data handling, nalysis and presentation of results are pointed out so the others can learn from past experiences. Problems of outliers and missing values have also been treated to some extent. Hypothetical data was generated in computer under different situations in order to show the type of pattern of residual plot that can be observed under failurity of different assumptions. Formal techniques used to test failurity of some assumptions has also been explained. Particularly, tests for homogeneity, additivity, normality, outliers and influence was introduced. To help readers apply some of these techniques to their data sample programs from INSTAT, SAS and GENSTAT software are presented.
Show more [+] Less [-]AGROVOC Keywords
Bibliographic information
This bibliographic record has been provided by Ethiopian Institute of Agricultural Research