Iteration process to determine the optimal log-linear decay curve

StepProcess for each subject
1Fit the full model; i.e., fit a linear regression (as defined by equation 3) to all m parasitemia values of subject i
2Fit two models; i.e., fit the linear regression model to m − 1 parasitemia values by (i) removing the first observation and (ii) removing the last observation
3Determine the best model of steps 2i and ii, defined as the model corresponding to the minimum overall model P value
4Of the best model defined in step 3, repeat steps 2 and 3 in an iterative process until a minimum of four observations have been taken
5Of the m − 3 best models selected per iteration (including the full model [step 1]), the optimal model is defined by the minimum overall model P value