Exploration and description

Let us look at the summarised statistics again (Stats → Summary Statistics → Summarized Contents of Variates ...) with these two extreme values removed.

With the large difference in variation among observations when comparing planting dates 1 and 2 with planting date 3 one might consider a transformation to logarithms. Such a transformation depends on means of observations being approximately proportional to the variances of the observations. Dividing the means in the above table by their corresponding variances we obtain ratios of 7.5, 11 and 8.5 for planting dates 1, 2 and 3, respectively. This suggests that a logarithmic transformation could be appropriate. Applying a transformation can complicate the presentation of results. Since analysis of variance is a robust technique we shall first see what happens if we proceed without transforming dry corm weight.