error bars on correlation Figure

Question

[kanwarjag]

I have a figure of correlation as shown @ https://drive.google.com/file/d/0B8-ZAuZe8jldQnEzcThkOXlkNWc/edit However reviewer has asked to add error bars. I browsed around and could not find a method by which we can plot error bars on line graph. It is a correlation of two numerical values and we cannot show SE of one sample.

Any suggestion please.

Thanks

[Joe]

You mean you want x and y error bars?

[kanwarjag]

yes I think it come from replicates SE

Answer 1

[Sergej Andrejev]

If you have time to learn a little bit of R programming language you can use ggplot2 library to add confidence interval to your scatter plot. Here is an example showing correlation of horse power with miles per gallon

library(ggplot2) 

ggplot(mtcars, aes(x=hp, y=mpg)) + 
  geom_point() + 
  geom_smooth(method="lm") + 
  theme_classic()

[kanwarjag]

Thanks for the information. I have one clarification Let us say gene A has three values in RT PCR - 10, 20, 22 while gene expression has three value 2,3, 5, so we when we calculate CI we will take all 6 values of RTPCR and GE into consideration?

[Sergej Andrejev]

Good question. I don't think I am capable of answering it though. Check this response on stackexchange for answer: http://stats.stackexchange.com/questions/101318/understanding-shape-and-calculation-of-confidence-bands-in-linear-regression

Answer 2

[theobroma22]

Are you positive the reviewer means to add error bars to the correlation plot? I think he means to add error bars to the qPCR bar plot on the right since these measurements are done in triplicates and you have error. Also, you should present the validation plot as qPCR as a bar plot with error bars and the RNA-seq as a line plot, or vice versa. This can be easily done in excel which you used already. This being said, I'm assuming you do not have replicates for the RNA-seq analysis. Hope this helps.

[WouterDeCoster]

you should present the validation plot as qPCR as a bar plot with error bars

More appropriate would be a violin plot or boxplot.

[kanwarjag]

The graph is on left side @ https://drive.google.com/file/d/0B8-ZAuZe8jldQnEzcThkOXlkNWc/edit and this is where reviewer is asking to add error bars. I do have replicates (3) in RNAseq. So this brings my concern how to collapse RNAseq (3 readings) and QPCR (3 readings) corresponding to each point on graph either for SE or for CI

Thanks

Answer 3

[theobroma22]

As mentioned above you want x and y error bars, which corresponds to the fold-change expression on that axis. So, the x direction error bars correspond to the rna-seq values and y direction to the qPCR values. So there is no need to collapse anything. This is quite simple to do. See the error bar tab/section for the plot in excel. You can even highlight the cells that have the error values to use for each point and its direction.