Hi all. I'm new to Graphpad Prism and after fiddling with the program, I was having some trouble. I have colocalisation data of multiple cells (ROIs) and I want to visualise these values (Pearson's correlation coefficient) in a bar graph with the mean etc. This is normally pretty straightforward, but the problem is that the Pearson's coefficient is not normally distributed; to find the mean value for multiple ROIs, the data needs to be transformed using Fisher transformation, averaged and then transformed back. The problem is that if I put in the Pearson values, Graphpad automatically averages them. So my question is: is there a way to let Graphpad use a different calculation method for creating the mean (in my case, transforming first, then averaging, then transforming back), or is there a way I can do this manually? (I could simply plot the Fisher transformed data, but these values are not intuitive/meaningful for the reader so I'm really trying to avoid this)
Alternatively, is there a different method that is commonly used to visualise quantified colocalisation that I'm missing? Any help would be much appreciated!
I understand your concern about visualizing non-normally distributed data in GraphPad Prism. While Prism doesn't have a built-in option to perform the Fisher transformation and average the values before plotting, you can work around this limitation by following these steps:
1. Perform the Fisher transformation on your Pearson's correlation coefficients using the formula:
z = (x - μ) / (σ / sqrt(n))
where x is the Pearson's correlation coefficient, μ is the population mean (which is zero for the Pearson's correlation coefficient), σ is the standard deviation, and n is the sample size (i.e., the number of ROIs). You can calculate these values separately for each ROI, and then apply the Fisher transformation. 2. Once you have the Fisher-transformed values, you can plot them in a bar graph using Prism's "Bar Graph" function. To do this, select the Fisher-transformed values, go to the "Graph" tab, and click on the "Bar Graph" button. 3. In the "Bar Graph" dialog box, choose the column containing the Fisher-transformed values as the "Values" column. If needed, adjust the other settings, such as adding error bars or modifying the appearance of the graph. 4. When you're done, click "OK" to create the graph. 5. Finally, to add a means line to the graph, follow these steps:
a. Select the entire dataset (including the Fisher-transformed values and any additional columns you may have). b. Go to the "Analyze" tab and click on the "Mean" button. c. In the "Mean" dialog box, select the column containing the Fisher-transformed values. d. Choose the "Standard Error" option under "Calculate." e. Click "OK." 6. A means line will now appear on the graph, representing the average of the Fisher-transformed values.
This approach should allow you to visualize your non-normally distributed colocalization data in a meaningful way. An alternative method would be to plot the raw Pearson's correlation coefficients directly, but as you mentioned, this might not be very informative for readers who aren't familiar with the scale. By applying the Fisher transformation, you can present the data on a more interpretable scale, making it easier for readers to understand the strength and direction of the relationships between variables.
Hi Saif, thanks for your quick and elaborate response! However, what I meant was that I wanted to plot the Pearson's coefficient values in a meaningful way, NOT the transformed data since those numbers are just arbitrary values to obtain a normalised population. The Pearson's scale would range from 1 (full colocalisation/dependency) to 0 (no correlation) to -1 (mutual exclusivity), whereas the Fisher transformed values have no meaning. I think I will just plot the spread of all values without a mean, and manually add the mean, p-value etc. to keep the intuitive visual and provide all relevant parameters.