I want to compare significant difference between two dose response variable. For example: Toxicity of a test item under two different temperature conditions.
Thanks for the answer. But my ? is how to analyse two dose respond variables (the treatment's response is either increasing or decreasing).
For example:
a test item's toxicity on an organism with two different temperature say, Mortality of X organism exposed to different concentrations of the test item. The mortality will be in dose respond manner. I want to see the difference between the two temp. overall
Do several assays for each of the two conditions, and for each one estimate the response (e.g. as an LC50 or slope). Then compare the means of both sets of measures by a method like t-test, Tukey, or randomization test (this is even better because you don't have to assume normality).
Thanks for the answers. But, the t test is not appropriate, because the treatment group has more than two means (Minimum 5 concentrations for each condition exposure). If ANOVA is performed, It will not give overall comparison between the two conditions.
My 2 cents: It would depend on how the experiment is done and what is you primary hypothesis. If the dose responses are collected from same experiment subject, then the responses are correlated repeated measures. If you are interested in LD50, you can get LD50 for each subject and simply use t-test to compare the difference in LD50. However, if you are interested in comparing dose-response profile between two groups, you can compare the slope difference by using group by dose interaction. Depends on whether it's repeated measures design or not, you can use mixed model or glm to analyze the data, and run post-hoc pair-wise comparisons to compare difference at each dose with correction for multiple comparisons.
Thanks for the answer Dr. Fang Young Li and Dr. Di Maria. My Null Hypothsis is... there may not be significant difference between two temperature conditions of an organism when exposed to series of test concentrations of a test item.
I have observed parameters such as mortality, weight change, reproduction ability. etc in all the concentrations as dose response manner. I want to see whether the two conditions are same in all the aspects or different. Overall comparison needed 9Not each dose level)
You have two factors i.e. temperature and series of test concentrations.
Your response variables are mortality, weight change, reproduction ability &
you want to see whether interaction of temperature and series of test concentrations have an influence on your observed parameters.
If I correctly understand you
You can use logistic regression with interaction term (interaction of temperature and series of test concentrations) for your observed parameter (mortality)
You can use two way anova or linear regression with interaction term for your observed parameter (weight change).
I agree with Samson's and Fang Yong's response, and would suggest using a mixed effects model, with subject as a random factor, this way you can test for everything else while controlling for measuring multiple aspects of the same subject. Hope that helps!
I don't use SAS, but I think you'd need to look into 'PROC GLIMMIX'. I use R, which is free. Lots of good resources for doing this there - for example see this tutorial: http://jaredknowles.com/journal/2013/11/25/getting-started-with-mixed-effect-models-in-r Its a learning curve to work with R, the price is always right. Good luck!
I know this is an old question, but I would only use a linear mixed effects model or a generalized linear mixed effects model if you assume there is linearity in the dose-response over time. If you are using several doses administered to the same subjects over time at different dose levels you should be ok with linear models. If you are sampling repeatedly from the time course of a single dose for each dose level, you may need a different model. As dose tends to be a curve over time determined by bioavailability, you may need to use nonlinear mixed effects modeling to look at dose over time with temperature included as a predictor. Population PK/PD modelling analysis methods would help a lot here if that is the case.