Note a minimum of 6 dose points is required to fit a 5PL model.
|
and
Where y is the response and x is the log (dose), for the reference standard and the test sample respectively.
 is proportional to the slope at the midpoint.
If > 0:  is the left asymptote
 is the right asymptote
If < 0: is the right asymptote
 is the left asymptote
B is proportional to the slope at the midpoint.
E represents the asymmetry, and is non-negative.
ε is an error term with distribution N(0,1) so σi ε is the error with distribution N(0, σi2), for i = Ref or Test.
|
|
QuBAS allows the user to select whether the left asymptote should be A or D, or to try both. The possible choices are shown below:
A left lower A left upper A right lower A right upper
When E = 1 the 5-PL model becomes the 4PL model. When E is not 1 the 5PL is asymmetric.
For E < 1 it approaches the A asymptote more quickly than the D asymptote; for E > 1 this is reversed. This is shown below.
Asymmetry
For the parallel fit, 
= 
, 
= 
and the variance, σ2, is common to all dose groups.
and 
Note that the C parameter is not the midpoint in a 5PL fit.
