cxli233 / friendsdontletfriends Goto Github PK

Also known as spider graphs.

In short:

• shape is too sensitive to the axis ordering
• plotting more than one series is not readable
• the area scaling is misleading

This is a summary of the demonstration made by dataviz in this article: https://www.data-to-viz.com/caveat/spider.html.

Hello. Is heatmap plotting to overcome outlier bias at 95th percentile visualization applicable to scRNA-seq heatmaps?

A common problem, in my field at least, is plotting multiple dependent measurement on each sample. For example, if a researcher applies a drug to a dish of cells, then measures 100 cells in that dish, they often then plot 100 datapoints. This is incorrect, because there was only one sample, and the repeated measurements on that sample should not count toward the sample size.

We wrote about it here: https://rupress.org/jcb/article/219/6/e202001064/151717/SuperPlots-Communicating-reproducibility-and

Howdy there,

This isn't an issue, so feel free to close it, but I wanted to say that I think this is a fantastic project!

Would you be okay if I were to recreate this project as a webpage using interactive Plotly charts (with credit, of course!)

Cheers!

Hi,
I went through my own collection of plotting "sins" and I wanted to suggest some inspiration:

Friends do not let friends use 3D plots.
this is common in business presentations were pie charts (ugh), bar charts or maybe the worst ind 3D scatter plots are shown in 3D and with some sort of perspective which of course means that plots that are "further away" are smaller

Friends do not let friends different colors for the same treatment.
I see this sometimes in plots generated by students, that they choose color palettes but then do not make sure that the colors are consistent in multiple graphs panels in the same figure.

Friends don't let friends use scatter plots for very dense datasets
I sometimes se scatter plots were thousands of dots overlap into a single mass. Instead hex-tile maps or contourline plots are much more readable.

Friends do not let friends use box plots for non-bell-curve-distributed datasets
I have seen some authos make the argument that box plots are generally a relict of the past (when graphs had to be hand drawn) but I think they have their place, However for some datasets they are just misleading.

Friends do not let Friends use bubble plots to show increase/decreases in values
There is an argument that bubble plots are generally problematic but they can be used well if combined with different colors (e.g. in a dot plot) BUT I have seen people using them to show e.g. increase in budget over multiple years and it is incredibly hard to see by how much an area has increasesd becaseu people automatically ompare the diameter of a circle and not the area.... a double sizes pizza does not look twice a big. Most area based plots suffer from this problem.

Friends do not let friends use 1000 colors in a single plot
Sometimes peopel forget to put thier data into relevant categories and instead give each an individual color. e.g. when they plot something for every country on earth. It is important to use relevant categories (e.g. continents) and keep the amount of colors low.

An explanation as to why these incredibly popular stacked bar charts should actually be heat maps?