statistical analysis, probabibility density function, negative log-likelihood values

good day all,

I had a wave data set of a length of 3000. then i sorted out all the peaks and valleys of this wave. and came out with 75 amplitudes which i put in a matrix:

amp=[0.1646 0.0829 0.1354 0.2488 0.0915 0.1415 0.1646 0.0805 0.1720 0.1878 0.1537 0.0988 0.3110 0.3720 0.1683 0.3829 0.2220 0.0402 0.1841 0.2732 0.1744 0.0829 0.2244 0.1427 0.0976 0.0902 0.2598 0.1671 0.1256 0.4280 0.1695 0.1720 0.1463 0.3720 0.1244 0.0939 0.0817 0.3988 0.0902 0.2268 0.2476 0.1500 0.2012 0.0622 0.0390 0.1573 0.0756 0.2744 0.2122 0.3573 0.1171 0.3378 0.2098 0.1317 0.2683 0.3146 0.1878 0.1537 0.0756 0.1829 0.3122 0.1232 0.1537 0.1902 0.2829 0.1427 0.4634 0.0829 0.1305 0.1476 0.0780 0.0549 0.3061 0.1524 0.2280];

after that, I chose Normal, gamma, lognormal and Weibull distribution function types to obtain their Negative log-likelihood values, which gave me

• paramGaussian=normfit(amp); NgtvLgLkhd(1,1)=normlike(paramGaussian,amp);

    paramGamma=gamfit(amp);
    NgtvLgLkhd(1,2)=gamlike(paramGamma,amp);
    paramLogn=lognfit(amp);
    NgtvLgLkhd(1,3)=lognlike(paramLogn,amp);
    paramWbl=wblfit(amp);
    NgtvLgLkhd(1,4)=wbllike(paramWbl,amp);

NgtvLgLkhd=[-46.9030 -75.8580 -75.2367 -76.2557];*

then, i plotted the histogram of amp and Weibull function fitting curve, the result went very well. So, my quenstion is, did this mean that amp value follow a unique weibull distribution? was it because the

magnitude of negative log-likelihood value? any further validation methods? Thanks. Cheers.

all the best

Will

Answer 1

[bym]

Not sure about the negative log likelihood, but you could try a non-parametric estimate like Anderson-Darling for further validation

[Will]

thanks, man. although i have no ideabout Anderson-darling.