Dear Dr.colleagues, I am currently plotting the dynamics of an infection in Excel. For each sampling day, I plotted the mean infection load of the population, using the standard error for the error bars. When representing the values within their natural scale, error bars are symmetric. However, if I use the logarithmic scale offered by in Excel, the error bars (the SEs) become asymmetric.
Looking for a solution, I have found the attached tutorial, but there are some confusing aspects for me that I would like to understand in depth. If I understood correctly, it first use the absolute error and not the SE (right?). Then, transform the absolute to relative error using the 0.434x(δy/y), where δy is the absolute error (still right?) and plot the value of log(y) with the error bars 0.434x(δy/y)
My questions/doubts are the following:
It would be very useful if you can recommend me any lecture or, better, a tutorial at this respect. Thanks in advance,
Mario
Dear Sir
You can reed this documents
https://faculty.washington.edu/stuve/log_error.pdf
You would make your life a lot easier if you'd simply log the data and work with this. You can calculate means, standard deviations, standard errors and confidence intervals using the (log-)transformed values.
Note that a difference in logs represents a realtive change in the original scale.
ad 1) Yes, because differences in logs represent relative changes in the original scale.
ad 2) It's difficult what SE means if the distribution is not symmetric. Better use confidence intervals, and here the confidence intervals should be about relative changes.
ad 3) That should not be required at all.
It's hard to find a good lecture about that, as this is stuff covered in school.
- Badges
- Science topic
- Similar topics
- Counseling Psychology
- Confusion