how to create a nice image with very different number like 22 and 30000 (qPCR results)

Question

[rivir]

I have aproblem to create anice graph for a paper with results like this:

      CACNA1C   TNNT2   MYL2    
EMF     29      296     66863
VT3     3.4     137     221

[Devon Ryan]

Try a log10 scale. If that doesn't work, you can always split the axis. That's assuming you really need to see the lower values and the upper one.

[dariober]

Since you have qPCR in the title, I suspect log scale is the way to go, as Devon suggests. However, to get better suggestions you should say what it is that you want to emphasise with these data.

Answer 1

[Ibrahim Tanyalcin]

Dear,

They have already mentioned log scale here, so you can first try that out. From my experience, human brains are not hardwired to immediately interpret magnitude of numbers so you can try expressing them in means of shapes. Try to express them in means of the volume of sphere/tesseract/cone/tethahedrals.

[John]

Ackchyually, hehe, the human brain is least capable of interpreting volumes of blobs. Even log-transformed positions are better than volumes, although I agree it's probably best to highlight the value of objects in a log-scale as extra redundant visual information.

Original article about it: https://www.cs.ubc.ca/~tmm/courses/cpsc533c-04-spr/readings/cleveland.pdf Very nice blog post about it which references the above: http://flowingdata.com/2010/03/20/graphical-perception-learn-the-fundamentals-first/

[Ibrahim Tanyalcin]

There are zillion of things that changed since 1984. There are 2D graph options which we are all familiar, and there are other options which I think researchers should tinker and experiment with.

In visualizaition of data, conveying the message you want to the reader with cost of distortion is way better than conveying the raw data with cost of obfuscation.

There so many rendering options you can try on 3D volumes if one is very concerned with the reader understanding the raw data behind. An extra dimension = extra information. If you cannot comprehend difference between 2 3D cubes, you can always rotate them around to yield squares.

Moreover, you can distort the 2D chart and layer them to create pseudo-3D charts where the information from 2D is arguably conserved with the addition of now you having extra layer of information to display.

None of these options existed by the time that paper you linked was published. You can add interactivity and mouse over 2D charts if the reader is interested in the raw data. But if you are interested in highlighting a message, there are many options available: 2D charts--->we all know, volumes are another solution researchers should try.

Lastly, tinkering does not hurt. This discussion if getting off topic, we should let rivir experiment and tinker and choose what fits him best.

[John]

I disagree, but you are right, it's off topic and other than the study I linked to and the blog post by Nathan Yau, I have little else to add (and I doubt i'd change your mind anyway).