A Probit Planes Method for Analyzing Seed Deterioration Data

Abstract

Efficient management of seed stocks depends on an accurate description of seed longevity in storage. This work was conducted to study the implementation of the probit analysis method of modeling seed deterioration. Samples of field bean seed ( Phaseolus vulgaris L.) were stored at eight temperatures (20‐60°C) and four moisture levels (12–24 g H 2M O kg −1 ) from 0 to 372 d, and evaluated using the standard germination test. Probit analysis was applied to the entire data set simultaneously using nonlinear regression as well as the conventional environment by environment approach. Nonlinear regression was performed on the germination angles using the model statement: germination angle = arcsin√Φ[ ki − p/10 (Ke−CwLogM−ChT−CqT2 )] where Φ is a function that computes the probability that a random variable with a normal (0,1) distribution falls below the argument of the function; Ki is a seedlot constant, and Ke, Cw, Ch , and Cq are species constants. Regression yielded the following parameter estimates: Ki = 6.68 (95.4% initial germination), Ke = 9.08, Cw = 5.20, Ch = 0.0057 and Cq = 0.00079. Use of these parameters resulted in a fit ( R 2 = 0.83) comparable to that obtained by the conventional environment by environment approach ( R 2 = 0.80). Persistent lack of fit of deterioration curves to the normal distribution model occurred in both approaches. When probit analysis was performed by environment, specifying the initial germination of the seedlot as the natural response rate, linearity of the probits over time was significantly improved. This suggests that probit analysis of seed deterioration should be performed specifying a natural response rate (6% initially dead seeds in this case) if the initial germination is below 100%