SCREEN PLOTS¶

LOG (ChiSq) vs. iteration number

This plot will appear after the second iteration of the MaxEnt routine. If the ChiSquared is nearly constant for 3 or more consecutive iterations, this plot will not appear. The “====” bar in the plot indicates the target value of “N”, the number of intensity points.

Entropy vs. iteration number

This plot will appear for every iteration after the second. The “====” bar in the plot indicates the entropy of a flat distribution with the same number of diametral bins as have been specified.

Residuals

The standardized residuals are the difference between the intensity that is calculated from the distribution and the input intensity, all divided by the input error. For the model to fit the data well, this plot should look featureless (a.k.a. random). The “====” bars are at +1 and -1 standard deviations. 67% of the points should fit within the bars. If there is some systematic difference beween the model and the data, the residuals will reveal it by showing some shape.

Distribution

The distribution plot appears at the end of each iteration and shows the most recent distribution, whose calculated intensity is to be compared with the input intensity and errors. The values in this plot are weighted (multiplied) by the bin width. This means that when the bins are distributed in a geometric series, it will be quite difficult for the user to see a small peak at smaller diameters in this plot. Have no fear though because this method weights the volume fraction in a manner equal to that of the algebraic series.

Volume weighted size dist.: V(r)N(r) versus r

Once the MaxEnt routine has decided that it has a solution, this plot will appear. The vertical scale is the volume distribution (technically the “volume-weighted differential number distribution”). This is almost the same value as was plotted in the “Distribution” plot except the bin width has been divided out. The horizontal scale is a linear axis on which is plotted particle radii.

Note

The MaxEnt routine does its work with respect to particle radii. All answers are properly scaled to diametral units in the output files. Additionally, the MaxEnt routine works with intensities in 1/m units. It seems to do some bad things when the intensities are in 1/cm. The FORTRAN code isolates the user from this eccentricacy. All unit conversions are corrected in the output data.