I'm not sure if this question entirely belongs here but it is about physics and informatics/data analysis.
I've graphed some data pertaining to an nMOS MOSFET device as to how the current varies with voltage. The equation describing this function is
I=(k/2)*((vgs-vto)^2)*(1+(lambda*vds)).
Now when i graphed this data using an excel type program i tried to fit a trend line to it. From the form of the equation i expected a power relationship to yield the best R^2 value but instead the best R^2 value came from a logarithmic trend line.

Question

I'm not sure if this question entirely belongs here but it is about physics and  informatics/data analysis.
  I've graphed some data pertaining to an nMOS MOSFET device as to how the current varies with voltage. The equation describing this function is 
I=(k/2)*((vgs-vto)^2)*(1+(lambda*vds)).
  Now when i graphed this data using an excel type program i tried to fit a trend line to it. From the form of the equation i expected a power relationship to yield the best R^2 value but instead the best R^2 value came from a logarithmic trend line.

anonymous · Answer

(I will need a trend line to deduce lambda.) So my question is; should I just accept a mediocre R^2 value with a power relationship curve or go with a logarithmic trendline which fits better yet that function doesn't even feature in the equation.