nlparci for log-transformed data

Hello,

I am using lsqcurvefit for data fitting of a dataset. I am log-transforming the data to estimate the parameters and am getting a good estimate of the values. But when I use "nlparci" to find the confidence intervals for these parameters, they are really large. "nlparci" gave reasonable confidence intervals when I did not log-transform the data. But I want to use log-transform to later implement log-additive error model. Is there other way to get the confidence interval?

This is what I am doing now:

[PK_params1,resnorm,residual,exitflag,output,lambda,J] = lsqcurvefit(@(PK_params1,xdata)LSQ_MulticompODE_Amount(PK_params1,R,xdatan),PK_paramsguess,xdatan,ydatan,lb,ub,options);

ci = nlparci(PK_params1,residual,'jacobian',J)

percent_error_estimate = ((PK_params1' - ci(:,1))./PK_params1').*100