3 replies to this topic

[frrd@lundbeck.com]

I have some activity data from mice in time bins. The observed data is in percent of time they are mooving. If the mice are stimulated many will move around 100% of the time. However, some of the mice that moove 100% of the time are more stimulated than others so we observe a ceiling effect. So the real Emax is higher than the observed Emax which is limited by a ceiling of the measurements. I would like to convert my PD model with a ceiling effect so that. eg (0,0.3,0.9,1.2,1.5) preddicted from the model is converted to (0.30,90,100,100). I tried

 

logit(Emax*E/EC50+E)

 

or 

 

E= ln (EB / (1 - EB))

EB= Emax*E/EC50+E

 

However both ln and logit does not work in the model structure. Is there another way to do this effect prediction with a ceiling? 

 

BR Frederik

[smouksassi1]

HI Frederik,

you might need a trauncated model where you have an upper limit of quantification here Emax

 

you will need to write your LL statment as we do not have this built in.

 

Bests,

 

Samer

[frrd@lundbeck.com]

ok. I will look into how to write my own LL statement. 

 

Thx Frederik

[smouksassi1]

also look at : https://support.cert...l=+upper +limit