When fitting a drm model to my dose response data the obtained ED50
is off
log(ED50)
, log2(ED50)
, log10(ED50)
# here is some actual data
test_response <- c(0,1.130912987,-2.0159852,6.158574058,5.679161469,12.97,20.46711451,35.55009039,66.31837697)
test_dose <- c(0,0.006103516,0.024414063,0.09765625,0.39,1.5625,6.25,25,100)
test_df <- data.frame(dose=test_dose, response=test_response)
# fitting a Hill formula
m1 <- drm(response ~ dose, data=test_df, fct=LL.4())
# optaining ED50
ED(m1, 50)
I get an estimated ED50
of 1649
. When I fit the same data in a proprietary program I get a similar plot but with an (correct) EC50
of 45.5
I tried log10 as well as log2
of the calculated ED50
but it's off anyways
Solution
So I figured it out, the models were correct, but apparently there's when using ED()
one can choose between type="absolute"
and type="relative"
where "relative" ist the default; in my case given those response values are absolute, changing this parameter gave me the right ED50s