Single terminal inflorescences from two-stemmed plants of Rhododendron L. cv. 'Prize' and 'Gloria' were recurrently sampled during a 9°C cold treatment and subsequent forcing period in an 18°C growth chamber until plants reached anthesis. At each sample period, bud length and width were measured on each inflorescence, while subtending bud scale, petal, stamen ('Gloria' only), style, peduncle and ovary length, ovary width, and flower weight were determined on the two most basally attached flowers within an inflorescence for each cultivar. Polynomial regression models were calculated for each floral growth measurement over time, except for bud length and width, for each of the four flowers within a cultivar. Comparison of regression models revealed differences in floral part growth patterns among the various flowers within each cultivar. During the first week of placement in 18°C forcing temperatures after the cold treatment, bud length was the only floral part to grow on 'Prize' plants, while only female flower parts grew on 'Gloria' plants. These results indicate differences in growth resumption patterns between cultivars after cold storage. However, petal and style length were strongly correlated to flower weight for both cultivars indicating that these measurements could be used in laboratory or field situations to determine the floral stage of growth. The growth equations described would be useful for controlling the variability among shoots and flowers for analytical determinations, such as hormonal content analysis, and for studying relationships of floral growth patterns to environmental stimuli that could be used to control flowering. The statistical techniques used in this research make it possible to reduce the number of plants for such studies, as measurements taken on different flowers on the same plant can be used for floral part growth comparisons despite the inherent correlation present between such measurements.Key words: morphology, dormancy, growth modeling, multivariate multiple regression, controlled environment.
Lack-of-fit Testing for Polynomial Regression Models Without Replications
