Dear Fellow Researchers,
We presented dose-response curves as supplementary information to a section of our study which has X-axis as Log[concentration] because the concentrations were graded exponentially so the model we used to generate the dose response curves did include Log[0] because Log[0] is not a defined value. A reviewer suggested that we include 0 concentration which we provided a rebuttal stating the facts for which the Log[0] was excluded by the model and as such the reviewer recommended to Editor that manuscript should not be accepted in that form.
Reviewer recommends that if we cannot include Log[0] then dose-response curve is meaningless. That we should remove the Log[concentration] and convert all back to ordinary concentrations on X-axis.
How do I fit exponential concentrations ranging from 0 to 3125 to give a Sigmoid curve on a linear scale, without manual improvisation? If concentrations were linear then that I think may be possible. Please clarify.
I have seen many dose-response curves in many peer-reviewed reputable journals including Nature, Science and Cell that provided dose-response curves with Log[concentration] and excluding Log[0]. How do you respond to this type of reviewer that kept insisting that we are wrong if we don't include 0 concentration on the curve? His reasons were that 0 is the control that will help to determined NOAEC since our article is on risk assessment?
Looking forward to your responses?
Think about the Benchmarck approach and other ideas in the link:
https://www.researchgate.net/post/Log_transformation_of_values_that_include_0_zero_for_statistical_analyses2
Dear Olushola Awoyemi,
In general sigmoidal .... is the model. That means you have chosen a model, and it means that the selection has been evaluated by parameters assessing the fit of the model (eg AIC, BIC etc). So you have appropriate justification of the model based on relevant criteria. If you decide that best model is for example sigmoid Emax model with a baseline effect parameter (WNL model 106) then 0 is not a defined value. In many studies nonzero baseline effect is present (heart rate for example). If it your case and in your model baseline nonzero effect exist it cant be stated that 0 is the control. 0 could be element of the model or not its depend on your model justification and selection process. It should be your argument for discussion with Editor.
Best regards
Tomasz
@Prof. Tomasz, thanks for your response. Exactly how I presented my rebuttal but the Reviewer seems to want to defend his point as well and recommended to the Editor not to accept which the Editor did rejected our work ignoring the positive recommendations of the other two Reviewers.
Editor said since the research is outside his expertise he had to defer to the Reviewer's comments and recommendations. Seems unfair to me!
I am planning on resubmitting back to the same Journal and requesting another Editor. Many of the Editors of the Jounral seems not to have background or expertise closely aligned to our work but there are many articles published in my field in the same Journal. So I wonder how that goes.
In between, are you a toxicologist or have experience with Toxicology studies maybe I can recommend you as a Potential reviewer? Thanks!
Regarding the model: The measurement for C=0 should be treated as any other data point. Otherwise you would discard information.
Regarding the graphical presentation: I would present two graphs:
1) E vs. C.
2) E vs. log(C). Since log(C) for C approaching zero is -inf. I would use two subplots (or broken X axis): one for C=0 and the second for the other concentrations.
- Badges
- Science topic
- Similar topics
- Chemistry
- Pharmacology