Datasets for Statistical Analysis

Number of Cyclamen flowers

Keywords:

flowers, cyclamen, Poisson

Categories:

biology;
generalized linear models, Tweedie models, regression

Description:

This data comes from an experiment on induction of flowering of cyclamen. Four varities of cyclamen were subject to a combination of 6 temperature regimens and 4 levels of fertilization. The temperature regimens are combinations of five day-time temperatures (14, 16, 18, 20 and 26 degrees C) and four night-time temperatures (14, 16, 18 and 20 degrees C). Not all the combinations of temperatures are present. The response is the number of flowers, which vary from 4 to 26 with mode 8.

Variables:

Variety	The variety of cyclamen
Regimem	Temperature regimen (combination of the day-time temperature and the night-time temperature)
Day	Day-time temperature (in degrees C)
Night	Day-time temperature (in degrees C)
Fertilizer	Level of fertilization
Flowers	The number of flowers

Data Quality:

There are no missing values.

Source:

OzDASL

Notes:

• The flowers counts look distinctly under-dispersed relative to the Poisson distribution. There appear to be two few small counts, so that the counts are right skew relative to the Poisson distribution.
• Can try to estimate a Tweedie power family variance. The MLE estimate for the variance power is 2.46 while the REML estimate is virtually the same at 2.47.
• There is a dispersion effect for Variety. Varieties 3 and 4 have dispersion about 1/3 larger than the other two varieties.

References:

Gordon Smyth has some notes on fitting models at OzDASL.