I'm trying to apply a non-linear regression model for protein binding in R.
Data are as follows:
data$WT
[1] 107.194364 95.986477 87.511449 74.028678 67.733609
[6] 52.117508 38.486519 24.197712 15.854248 8.641564
[11] 5.965327 2.871084
data$So
[1] 2.0000000 1.0000000 0.5000000 0.2500000 0.2000000 0.1000000
[7] 0.0500000 0.0250000 0.0125000 0.0062500 0.0031250 0.0015625
Model is:
bindmod <- nls(WT ~
(Ase* ((Kd+So+0.03)- sqrt((Kd+So+0.03)^2 - 4*So*0.03)/2)/So)
, data = data, start = list(Kd = 0.03, Ase = 2000))
On trying to fit this error appears:
Error in numericDeriv(form[[3L]], names(ind), env) : Missing value or an infinity produced when evaluating the model In addition: Warning message: In sqrt((Kd + So + 0.03)^2 - 4 * So * 0.03) : NaNs produced
The value of this term within the square root should not be negative with the values present, however, so I don't understand why it is producing NaNs. Is this something to do with the estimation process?
Any help would be much appreciated, thank you.
George
You are not specifying the lower/upper bounds, so when kd and soare so that (Kd+So+0.03)^2 - 4*So*0.03 is being negative, the sqrt produces NaNs. Try setting both the lower/upperbounds _and_ thealgorithm`:
nls(WT ~ (Ase* ((Kd+So+0.03) - sqrt((Kd+So+0.03)^2 - 4*So*0.03) / 2) / So ),
data = data,
start = list(Kd = 0.03, Ase = 2000),
lower = list(kd=0.01, Ase= 0.01 ),
algorithm = "port")