Hi everyone, My questions to the forum are: 1. I have fit the model to the data and by visual analysis the fit looks good. But I am not able to figure out why the final parameters are associated with high CV%? 2. In the plot time 0 is missing although in the Summary table the predicted and observed conc. At time zero is zero? Please find my Phoenix file as attachment. Any comments are highly appreciated. Thanks, Chetan [file name=QA_pharsight_2-20120822.phxproj size=544150]Certara | Drug Development Solutions (531 KB)
Hi Chetan, Second question first, you don’t specifiy which object or which plot you are referring to but I guess you mean “Observed Y and Predicted Y vs X”, which has Y on a log scale, so Zero can not be plotted, turn it linear and you will see it; 
Now to the model fitting question; 1. I have fit the model to the data and by visual analysis the fit looks good. But I am not able to figure out why the final parameters are associated with high CV%? You have 8 samples for a model of 5 parameters, only three samples around the terminal phase, and none in the absorption period. also for Sc 61 the last 3 points are also difficult to fit to elimination. SC 61 0 0 SC 61 0.5 179533.33 SC 61 0.75 161250 SC 61 1 89700 SC 61 2 12256.667 SC 61 8 405.66667 SC 61 12 60.766667 SC 61 24 119.13333 SC 99 0 0 SC 99 0.5 211916.67 SC 99 0.75 128875 SC 99 1 72266.667 SC 99 2 15360 SC 99 8 262.66667 SC 99 12 190.5 SC 99 24 157.16667 You may need to explore other models
Hello Simon, Thanks for your reply. As I have no points in the absorption phase, I was thinking of fixing the Ka and estimate other parameters. Can you guide what steps should I take or is there any other better alternative to tackle this problem ? Thanks and regards, Chetan
Hello Simon, I tried using the Phoenix model which gives an option to freeze the parameter. I did for three different values of Ka (5, 10 and 20). The S.Error and CV% for other parameters are small. For selecting the best among these three, should I look at the AIC, BIC and IRES to find out the best choice? Thanks and regards, Chetan [file name=QA_pharsight_2-20120825.phxproj size=839058]Certara | Drug Development Solutions (819 KB)
Hi Chetan! [quote]I tried using the Phoenix model which gives an option to freeze the parameter. I did for three different values of Ka (5, 10 and 20). The S.Error and CV% for other parameters are small. For selecting the best among these three, should I look at the AIC, BIC and IRES to find out the best choice?[/quote] I’m afraid freezing parameters doesn’t do you any good. How can you “know” that they reflect the “true” values? You simply don’t have enough data to support your model. BTW, a ka of 20 means an absorption half-life of 2 minutes. Which kind of wonder-drug are you working with? [color=#0000DD]“The combination of some data and an aching desire for an answer does not ensure that a reasonable answer can be extracted from a given body of data.”[/color] (John W. Tukey)