allin1 mode

If more than one preliminary MEM run is needed to obtain converged error estimates, use allin1 mode. Here is a sample call to MemExp in allin1 mode:

memexp.exe allin1 2 invert.no.err.def data.001 0.0 analyze.def err1.dat invert.err.def err2.dat err3.dat

At least nine arguments follow the progam name on the command line when MemExp is run in allin1 mode. The first argument specifies the mode and is set to allin1. The second is an integer $\ge$ 1, MXITER, which specifies the number of preliminary MEM-only iterations to be done to estimate the data's standard errors in the mean. MemExp performs a total of MXITER $+$ 1 inversions, all but the first inversion using time-dependent standard errors estimated from the previous inversion. The third argument, invert.no.err.def, is the name of the input file containing parameters to be used in the first inversion of the data using uniform error bars. The fourth, data.001, is the name of the data file to be inverted and may include the path of the file (the subdirectory where it is located). The fifth argument is a real number, D0, the data's normalization constant (equations 1, 2, 10, and 11). In the sample command line above, D0 is set to zero, which directs MemExp to estimate the normalization directly from the kinetics data. (Note that this automatic estimation is not recommended for data that have appreciable slope at short times when plotted versus log time. For such data, D0 should be set on the command line to a larger value and LAGMUL should be set to Chi2. Severely underestimating D0 can induce spurious maxima in f [10].) The sixth argument, analyze.def, is the name of the input file containing parameters to be used in analyzing each of the MEM distributions automatically chosen by MemExp. This analysis involves writing the MEM fit and residuals to a file and revising the error estimates to be used for the next numerical inversion. The seventh argument, err1.dat, is the name of an output file to which the kinetics data and the first estimates of time-dependent errors are written in three-column ( $t_i,y_i,\sigma_i$) format. The eighth argument, invert.err.def, is the name of the input file containing parameters to be used in all inversions of the data using time-dependent error bars. The ninth and any additional parameters each specify the name of a data file to be written with revised error estimates. The number of such parameters depends on MXITER. In the example above, MXITER = 2; so three inversions are done and the data files with errors derived from the second and third inversions are written to files named err2.dat and err3.dat, respectively.

If WRNORM = T (true) in allin1 mode, the data's normalization, D0, is iteratively refined along with the error estimates: D0 is adjusted for use in the next MEM inversion according to the normalization of the distribution recommended by MemExp in the current inversion. When the normalization D0 is known (or not to be changed by the program), set WRNORM to F (false) to avoid this adjustment of D0 in allin1 mode.