Tell me more ×
Cross Validated is a question and answer site for statisticians, data analysts, data miners and data visualization experts. It's 100% free, no registration required.
up vote 0 down vote favorite

I need to choose a non linear regression model for this kind of data

The experiment was a time course, for a drug, so I get phosphorylation response behaviour in time

In enzyme kinetics or a binding non linear regression model, you do not take into account time, instead you plot the substrate concentration Vs Vmax or radio ligand concentration (X) Vs total binding (Y). Here I got time (X) Vs kinase phosphorylation (Y)

Could you help me to get a proper model to do the fit?

share|improve this question
What makes this a non linear model? – Michael Chernick May 18 '12 at 6:21
The line you see is not fitted. The data gives you a similar graph like michaellis menten, but in that model there is no time involved, here I have response ~ time not velocity ~ [substrate concentration] – friveroll May 18 '12 at 6:40
With a non linear regression michaellis menten model, you could predict variables like Km and Vmax, so I like to find a model to predict similar variables. – friveroll May 18 '12 at 6:46

1 Answer

up vote 0 down vote accepted

This is repeated measures anova. The repeated measures are made at time points. It is anova because you are comparing groups. It is not true that the explanatory variable in regression cannot be time.

share|improve this answer
Yes I do ANOVA, for compare both groups, but I need to get a model for non linear regression, to get the fitted curves you see in the graph. – friveroll May 18 '12 at 6:28
For enzyme kinetics, there is a non linear regression model, Y = Vmax*X/(Km + X) but X is for substrate concentration. But here I am dealing with time, not substrate concentration. – friveroll May 18 '12 at 6:35
Thanks for showing me the nonlinear model. Repeated measures ANOVA is different fron ordinary one way ANOVA because of the special covariance structure so you might want to use proc mixed rather than proc glm or proc anova for that. . – Michael Chernick May 18 '12 at 11:09
SAS has a nonlinear regression procedure called proc nlin. You should be able to use that. Since you know the form of the model it shouldn't be hard to setit up with the appropriate input. Again there is no reason why time can't be a covariate for the model you use in SAS. You just would want to feel that an additive independent noise component is realistic. If the error term is very non normal you can use bootstrap. I don't know a simple way to guide you through doing a bootstrap on nonlinear regression in SAS. It might require programming your own macro – Michael Chernick May 18 '12 at 11:09
The data was analized with graphpad prism for one way anova, with dunnet post test for each group, and 2 way anova to compare between both groups. I really apreciate your suggestion for repeated mesures, I see an improve in p-values for comparisions. But may be I was little confused, yesterday. In this model Y = Vmax*X/(Km + X) you take into account substrate concentration, instead I have protein phosphorylaiton fold change, So the question should be if I can use this kind of model for my data. – friveroll May 18 '12 at 18:03
show 3 more comments

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.