weighted fit to log-log data

Hi, I need to perform a first order log-log fit with the ability to weight the samples. This should fit a straight line through data plotted on log-log paper. My code seems right to me but the results are wrong. Can any one set me straight?

x = independand data

y = dependant data

w = weighting scheme

d = lscov( [ones(size(x')) log10(x)'] ,log10(y)',w)

%if i plot this the fit line is out by a factor of ~10^3

loglog(10.^x,10.^y,'o',10.^[.2,.5],10^(d(1)).*10.^([.2,.5]*d(2)),'-')

[Ben]

Further to this, it seems to me that lscov is correct but my plotting is incorrect. This seems closer, but still incorrect:

loglog(x,y,'o',[.2,.5],10^(d(1))*10.^(d(2)*[.2,.5]),'-')

Any ideas where I'm going wrong?