The function multiplying parameter β4 is a steep sigmoid (or Hill function) whose value is close to 1 when y is reasonably larger than 0.
To convert parameters to molar units, it is necessary to multiply parameters by 10-9, whicorrespondsnds to a spherical cell of diameter 1. 5 μm.
First, we perturb one parameter of by multiplying the parameter value by either 0.99 or 1.01 to generate a starting point, Θ 0), for GD.
For a selection of the next set of the rate parameters, lognormal proposal distributions were used based on logarithm of random values selected from a normal distribution with mean 0 and standard deviation 0.05; this can be achieved either by adding a normal deviate to the log of the parameter, or by multiplying the parameter by the exponential of the normal deviate.
Because the optical signal is inversed with respect to the neuronal activity, the sign in the model has been corrected by multiplying all parameters by −1 prior to creating the images [14], [17], [27], [28]; hence darker regions would be expected to have the highest rate of neuronal activity.
We modelled this effect in a phenomenologic way by multiplying toxicity parameters with a factor ≥ 1 at time of first application of chemotherapy.
The indirect effect is found by multiplying the parameters along each of the (directed) paths from X to Y that include M and summing them; here, this is (α xβ m + γ xα lβ m).
To ensure that a global optimum was found, the fitting procedure was initialized with 5000 different parameter combinations derived from the manual estimate by multiplying each parameter with a random number between 10−3 and 103.
This was done simply by multiplying the parameter estimates (Beta) obtained from the GLM with the surface (S) under their respective predictor (S = BetaauditoryxSauditory + BetaactionxSaction). Brain activity across conditions were compared using this surface (for details please see [4]).
Percentage signal change within the ROI was calculated by multiplying the parameter estimate by the peak-to-peak height of the EV model, and dividing by the mean signal across the whole acquisition period (Mumford 2007).
