0

I would like to know if the parameters of my model are identifiable. Does a high correlation between parameters, which are shown as a result of fit, a sign of parameters unidentifiability?

[[Variables]]
    alpha:    3.63101902 +/- 0.42188236 (11.62%) (init = 5)
    beta:     2.79949195 +/- 0.54011517 (19.29%) (init = 4)
    l:        63.3138794 +/- 10.1429942 (16.02%) (init = 90)
    yoffset:  31.7523156 +/- 34.5025649 (108.66%) (init = 0.001902903)
[[Correlations]] (unreported correlations are < 0.100)
    C(beta, l)        = 0.999
    C(alpha, l)       = 0.987
    C(alpha, beta)    = 0.981
    C(beta, yoffset)  = -0.762
    C(l, yoffset)     = -0.735
    C(alpha, yoffset) = -0.625

I don't know how to interpret these correlations, I'd be glad for your help!

qbush
  • 712
  • 2
  • 9
  • 28
Alicja
  • 1
  • 1
  • I don't know what you mean by "identifiable". The correlations are very high, but the uncertainties in the parameters seem "large-ish but not infinitely high". Of course, posting a complete but minimal example would almost certainly help understand what you are trying to do. – M Newville Dec 17 '22 at 02:53

0 Answers0