Hi, I'm using a non linear model to fit experimental survival curves. This model describes the fraction of "still active" experiments as a function of time t as follows: f(t)=(1+exp(-etaD*cD)) / (1+exp(etaD(t-cD))) Moreover, when experiments are still active, they may change of state (from 0 to 1). But they may fall inactive before changing their state (their state still equals 0). The survival curve of state may also be fitted with the following model: f(A)=(1+exp(-eta1*c1)) / (1+exp(eta1(t-c1))) * (1+exp(-etaD*cD)) / (1+exp(etaD(t-cD))) I estimate with nlm 1?) values of etaD and cD parameters and 2?) inject them as constant in the function to be minimized by nlm to estimate values of eta1 and c1. I perform these estimations for two different experimental conditions that both have their values of etaD,eta1, cD and c1. I would like to know if there is any statistical method to compare the estimated values of parameters of the two distributions ? And wether it's the case, how to perform it in R ? Hope I'm clear enough for getting help, Etienne ------------------------------------------------------------------- Etienne Toffin, PhD Student Unit of Social Ecology Universit? Libre de Bruxelles, CP 231 Boulevard du Triomphe B-1050 Brussels Belgium Tel: +32(0)2/650.55.30 Fax: +32(0)2/650.57.67 http://www.ulb.ac.be/sciences/use/toffin.html
How to compare parameters of non linear fitting curves
3 messages · Etienne Toffin, Spencer Graves
Is your first model a special case of the second with eta1 = 0?
If yes, what about using 2*log(likelihood ratio) being approximately
chi-square?
Can you recast the problem to use "nls"? If yes, might the 'nls'
methods for 'anova' or 'profile' give you what you want?
Hope this helps,
Spencer
Etienne Toffin wrote:
Hi, I'm using a non linear model to fit experimental survival curves. This model describes the fraction of "still active" experiments as a function of time t as follows: f(t)=(1+exp(-etaD*cD)) / (1+exp(etaD(t-cD))) Moreover, when experiments are still active, they may change of state (from 0 to 1). But they may fall inactive before changing their state (their state still equals 0). The survival curve of state may also be fitted with the following model: f(A)=(1+exp(-eta1*c1)) / (1+exp(eta1(t-c1))) * (1+exp(-etaD*cD)) / (1+exp(etaD(t-cD))) I estimate with nlm 1?) values of etaD and cD parameters and 2?) inject them as constant in the function to be minimized by nlm to estimate values of eta1 and c1. I perform these estimations for two different experimental conditions that both have their values of etaD,eta1, cD and c1. I would like to know if there is any statistical method to compare the estimated values of parameters of the two distributions ? And wether it's the case, how to perform it in R ? Hope I'm clear enough for getting help, Etienne ------------------------------------------------------------------- Etienne Toffin, PhD Student Unit of Social Ecology Universit? Libre de Bruxelles, CP 231 Boulevard du Triomphe B-1050 Brussels Belgium Tel: +32(0)2/650.55.30 Fax: +32(0)2/650.57.67 http://www.ulb.ac.be/sciences/use/toffin.html
______________________________________________ R-help at r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
Spencer, Le 22-avr.-09 ? 03:33, spencerg a ?crit :
Is your first model a special case of the second with eta1 = 0? If yes, what about using 2*log(likelihood ratio) being approximately chi-square?
Yes, the first model is a special case of the second with eta1=0? Could you give me more explanation about this method with likelihood ratio and chi-square ?
Can you recast the problem to use "nls"? If yes, might the 'nls' methods for 'anova' or 'profile' give you what you want?
I tried fitting the distribution using nls (perfect fitting). However, if I well understand 'anova' is used here to test alternative fittings (of alternative models) on a same distribution. Here I want to test the difference of parameter values between two different distributions and one same model (i.e. I want to test if eta1(TREATMENT) is different Do you have any other idea ? Thanks, Etienne One short question about nls: are there any reason why nlm should be used rather than nls and vice-versa (nls results are quite more full than those of nlm)?
Hope this helps,
Spencer
Etienne Toffin wrote:
Hi, I'm using a non linear model to fit experimental survival curves. This model describes the fraction of "still active" experiments as a function of time t as follows: f(t)=(1+exp(-etaD*cD)) / (1+exp(etaD(t-cD))) Moreover, when experiments are still active, they may change of state (from 0 to 1). But they may fall inactive before changing their state (their state still equals 0). The survival curve of state may also be fitted with the following model: f(A)=(1+exp(-eta1*c1)) / (1+exp(eta1(t-c1))) * (1+exp(-etaD*cD)) / (1+exp(etaD(t-cD))) I estimate with nlm 1?) values of etaD and cD parameters and 2?) inject them as constant in the function to be minimized by nlm to estimate values of eta1 and c1. I perform these estimations for two different experimental conditions that both have their values of etaD,eta1, cD and c1. I would like to know if there is any statistical method to compare the estimated values of parameters of the two distributions ? And wether it's the case, how to perform it in R ? Hope I'm clear enough for getting help, Etienne ------------------------------------------------------------------- Etienne Toffin, PhD Student Unit of Social Ecology Universit? Libre de Bruxelles, CP 231 Boulevard du Triomphe B-1050 Brussels Belgium Tel: +32(0)2/650.55.30 Fax: +32(0)2/650.57.67 http://www.ulb.ac.be/sciences/use/toffin.html
______________________________________________ R-help at r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.