Hi all,
I have generated two gene expression curves by measuring bioluminescence every 10 mins for 12 hrs. These curves tend to overlap at at some time points, while a difference can be seen at other time points. What is the best way to statistically analyze the curves and differences (if any)?
Thanks,
If you have two single curves, these are just two single curves. They may be more or less similar - but there is no way to make any statsistical inference since you don't have any information about the characteristics and the amount of dissimilarity one can expect.
If you have a sample of several (>2) curves that were measured under different conditions, one might be able to statistically analyze the correlation of some characteristic with the conditions. It typically remains difficult to specify the interesting (and meaningful) characteristic of the curves and how to extract/quantify it from the data. One is lucky when there is a theory behind suggesting a particular functional relationship, and it becomes more messy when the curves can only be described by over-parametrized models (such as splines). Because all this may not be trivial, I suggest cooperation with a local statistician.
This analysis is simple. It is good to use GraphPad Prism software. I have sent the pictures.
Abolfazl Ghoodjani , no you are wrong. It's not that simple.
Note that you checked "Consider each replicate Y value as an individual point" - this is surely NOT correct here. The points within each curve are not "individual", they belong to the same curve, obtained from the same individual/subject/specimen. I don't perfeclty understand the only other option "Only consider the mean value of each point" but I am sure that this is also wrong.
Apart from this, the analysis you proposed compeare slopes of regression lines. I am pretty sure that the curves Kiran was talking about are not just simple straight lines!
Hi all, I agree with Jochen Wilhelm . I am not sure about this suggestion but is it possible to treat these curves as a growth curve comparison?. I note that R has nice software for doing this. I would start the hunt here:
https://www.google.com/search?q=comparing+growth+curves&rlz=1C1CHBF_enUS917US917&oq=Comparing+Growrh&aqs=chrome.2.69i57j0i13l6j0i13i30l3.15258j1j7&sourceid=chrome&ie=UTF-8
Add R to the query to check on packages. Best wishes to all and good hunting, David Booth
Y = f1 (X) vs Y = f2 (X)
H0 : f1 = f2 vs H1 : f1 ≠ f2
The f function can be linear or nonlinear.
- If f is linear, we compare the regression coefficients and obtain the probability value.
- If f is nonlinear, then we better compare the curves and get the probability value.
For both methods, Prism software has a solution that I have sent pictures of.
and at the end
Jochen Wilhelm and David Eugene Booth are my teachers and I am their student.
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics