Curves not fitting the data well

[PhilPalmer]

Problem

The estimated gradient of the curve (parameter b) is too influenced by non-responders when fitting noisy data.

Implementation

• Update results
  • Add an option to view each sample's data and curve separately would make it easier to assess the fit of curves
  • Fit data to just the responders to estimate the correct gradient and then fit curves to all the data but with the gradient fixed to the responder estimate*
  • Report all model parameters (not just the IC50s)
• Update documentation

*Alternatively, set limits on the gradient before running the model. However, when David tried that with the data set which the problem was observed for the gradient estimate was just the lower limit.