I have data with some extreme values. I really like bar plots (not box plots) and wonder if its ok to present bar plots of median +/- mad?
However, when log-transformed the data is normal, but I don't want to plot transformed data. I want to show the real data and the variance.
Can I run statistics on the log-transformed data using parametric tests?
The results would be a bar plot with median +/- mad, with a "*" indicating significance based on parametric tests on log-transformed data.
Any experience?
Hi lee,
if you are interested to present to other than consider other technique which fits up the visualization of data. there is no problem with bar plotting. if you have huge dataset then it looks awkward while plotting and presenting people or audience.
yeah you can go up with logarithmic and it is very much considerable and easily understandable.
if you transformed data to logarithmic then it is not good choice to run again statitics!
you are welcome if you will have any questions.
Regards
Muddsair
Thank you both!
Im actually looking at other ways than bar plots due to your answers. Even though bar plots are the most seen in medical journals.
Im not going to transform my data and Im not going to run statistics on transformed data neither.
I figured I make box plots with individual values with lines drawn between conditions, then one can visualise up/down regulation individually. Then mark the median as well.
Finally, Im running non-parametric tests in every comparison.
Thank you again!