Hi all,
I'm trying to use the betamix function (betareg package) to create a beta
mixture model but get this error that I don't understand:
"Error in lm.wfit(x, linkfun(y), weights, offset = offset[[1L]]) :
NA/NaN/Inf in 'y'
Error in lm.wfit(x, linkfun(y), weights, offset = offset[[1L]]) :
NA/NaN/Inf in 'y'
Error in lm.wfit(x, linkfun(y), weights, offset = offset[[1L]]) :
NA/NaN/Inf in 'y'
Error in flexmix::stepFlexmix(fullformula, data = mf, k = k, nrep = nstart, :
no convergence to a suitable mixture"
The code I wrote was:
"beta_mix_model <- betamix(formula = def_relat ~ data_roads,
k = 2)"
I'm making the hypothesis that def_relat (ie the level of deforestation) is
a function of data_roads (ie distance to roads). When a pixel is far from a
road, its probability of deforestation is very close to zero with very low
uncertainty. Closer to roads however, deforestation is possible (with very high
uncertainty of the intensity of deforestation) but occurences of deforestation
also occur. This is why I'm considering to groups of observations, with a
beta distribution of probability for modelling a probability of deforestation
between 0 and 1.
I don't understand the error message, because none of the data I'm using
includes NA or Inf. Does anyone has an idea of the reason why I get this error,
or can tell me if I'm using the function betamix the wrong way?
Thanks,
Camille
[[alternative HTML version deleted]]