2.4. Growth and reduction signal modeling
Modeling was done using R packages NLS for parameter estimations
and NLStools for parameter initial value choice and
graphic output. All the processes were embedded in a script to
automate the treatment of raw data. Baranyi model (Baranyi and
Roberts, 1994), the modified Gompertz model (Zwietering et al.,
1990) and the Buchanan model (Buchanan et al., 1997) were
initially fitted. As the growth curve signals after the exponential
growth phase were jagged and led to mismatch for modeling, we
disregarded this portion to improve fitting. In order to determine
the best fit, the Akaike Information Criterion (AIC) was compared
between the models. Results obtained led to identification of the
modified Gompertz model as the best fit. Apparent lag (l*) and
apparent growth rate (m*) estimations were obtained by
modeling.