geopsy.org

Hi Marc,

When running dinver module, the misfit value doesn't exceed 0,0005.
The comparison between target (dispersion curve) and results of inversion let me think there's a problem of configuration of my runs, but which one...

Is misfit value equivalent to a RMS ?

I've got a practical question too about geopsy : How can I get error bars on my dispersion curve ?

Thanks.

The misfit is a RMS normalized by standard deviations if there are provided or by the values themselves if no error estimate is given.

I guess your dispersion curves comes from active source experiment. In this case there is no automatic definition of the errors like for array processing of ambient vibrations. If you don't know the error bars from experiment processing it is not necessary to invent something to provide to dinver. You can use the misfit computed without error bars as it is: a maximum of 0.05 means that the dispersion curve is adjusted with an approximate error of 5%. You have to use the "Minimum misfit" feature to really certify that no curve will exceed 5%. A minimum misfit specification has been recently introduced in dinver. It prevents individual misfits of each sample to be very small (below a reasonable precision). When you have very small misfits, you can perfectly adjust some parts of the curve but not very well others. With minimum misfit option, all produced curves are following acceptable shapes. One drawback is that minimum misfit makes the inversion a bit more difficult (a lot more strictly equivalent solutions).