Displaying a panda's family tree

Question

I have a section on my website that displays information about all known captive pandas, (Nearly 700) in total.

Each panda has a single page which display it's name, birth / death and where it has been located. The page also has a link to it's parents and it's children.

The main page(s) shows a table of all the pandas ordered in descending date or birth. I would like this page to display related pandas together.

This seems to imply some sort of family tree structure but for many families. The problem being that pandas mate with different pandas. so a female with 7 cubs could have 5 different fathers, who have cubs with other females, this leads to a very messy tree.

Does any one have any suggestions on how this could be displayed?

Answer 1

You have a nice data visualization challenge. Just throwing out this idea.

What if you don't automatically show the family connections on the main page. I suggest an interactive data visualization.

When you rollover a panda, then the 'connections' show up. The relationships can be shown with lines and/or with color coding, for example, to show mother/father/sibling. You can also reorganize the pandas to bring the 'family' pictures to another visual plane above the main page so that it may be easier to see the family together as opposed to scattered all over the screen.

The tricky bit is whether you want to show siblings related via mother or siblings related via father. Perhaps this can be a toggle that the user sets.

Ex:

• Toggle is set to showing siblings related via mother
• Rollover a panda to see its family (this panda gets a special highligh, let's say yellow)
• Upon rollover, it's family rearranges and shows nicely as a tree as described below connected by lines...Mother is at the top of the tree.
• Mother shown outlined in pink
• Siblings shown outlined in green
• Father of each sibling shown outlined in blue above each sibling and of the current panda
• You need a way to close out this family view...when the user closes this view, the pandas return to their original position on the main screen.

Answer 2

The "family tree" structure of pandas that you're trying to represent is a directed acyclic graph. It is not a tree. This cannot always be represented in 2D without intersecting lines. With 700 nodes and 2 ancestors each, it is likely to be messy.

Graph visualization tools can help with this problem. Start with a simple webapp to get an idea (put all the mother->child and father->child pairs into the list).

More advance programs such as yFiles and GraphViz are suitable for complex and large graphs. See gallery to get an idea about possible results.

Another approach is to try to simplify the problem, for instance by having ancestor/descendant trees focused on each particular individual.

Answer 3

A cub can only have one mother and father, why do the relationships have to be grouped by "family"? Seems like you should just group them by DNA, not by mother/father pairs.

---

UPDATE:

Group by (Father <> Child <> Mother) NOT (Father/Mother = Child_A, Child_B, etc.)

Just trace who's related to who by mother and father, don't keep track of who the brothers and sisters are... unless it's for mating reason. If you're comparing then for mating reasons to the general populace, just give a percent, or fitness rating as it relates to genetic overlap.

UPDATE 2:

Andi, is this the Panda index page:

• http://pandanews.org/the-pandas/

If so, how to you plan to group Panda's where one or more of the parents are "Unknown."