My lab has started producing time series Tumor Growth Delay graphs and we are running into some issues. We take the normalized average of tumor volume per week of mice in different treatment groups and produce a line graph. However there has been debates on how to handle mice which have their tumors completely regress and mice which have died over the course of the experiment.
In general how do we reconcile these events?
In this particular situation, data size is a crucial factor. If the data size is CONSIDERABLY large compared to these events, I would recommend to follow the very basic data cleaning option: truncate the data set by removing the missing events.
Can you share the summary of the data?
Article Growth Rate Analysis and Efficient Experimental Design for T...
Hi Keith,
There is something called : sample selection bias. For an experiment (or regression analysis) to be representative of the population, independence in your sample is desirable. If you remove some data from your experiment or regression, because some patients or mice dropped the treatment, your results may be biased. A traditional correction for sample selction bias was introduced by Heckman in his famous paper in 1979.
Please do not plainly remove from your analysis those mice that died, it may produce a bias in your results.
Best
Pablo
- Badges
- Science topic
- Similar topics
- Biological Science
- Kinesiology
- Running