This research was conducted to compare the suitability of logistic, segmented and linear regression models in estimating rate and duration of grain filling of three Iranian landraces and five cultivars of rice in two planting patterns of 15 ?15 and 30 ? 30 cm. The experimental design was a 2 ? 8 factorial in randomized complete block layout with three replications. The experiment was conducted at Rasht Rice Research Institute, in 2001. According to the linear and logistic models, low yielding landraces had lower rate of grain filling and shorter grain filling duration than high yielding cultivars. However, adverse results were obtained by fitting the segmented regression model. Also, in fitting the linear model the lag period of grain filling duration was longer for landraces. The direct effects of rates and duration of grain filling estimated based on the three models, on grain yield were positive in both planting patterns as well as their combined data. According to the results obtained, in conditions similar to those in this study, the linear and logistic models are suitable in estimating rate and duration of grain filling in rice.
)