On testing normality using several samples: an analysis of peanut aflatoxin data

Eight samples of size 16 and three samples of size 15 consisting of replicate determinations of aflatoxin in peanut subsamples were considered. These data were analyzed to decide whether the samples could have arisen from two-parameter normal or lognormal parents. The method of analysis consisted of first transforming the individual samples in such a way that the transformed values had uniform distributions on the unit interval under the null hypothesis. The transformed values were then pooled and tested for uniformity. Results of the analysis indicated that the normal model fitted the data quite well and definitely better than the lognormal model.