Long train of derivations

Cryptoref, it seems to me that what you are trying to cite (with the exception of the original DNA data from AncestryDNA) is not a source or a series of sources, but a process and a set of tools you have used.

If this were my citation I'd focus it on the original AncestryDNA "report." Then I'd explain, free form, the process and the tools that led to thus-and-such conclusion.

Cryptoref, you write:

"There is this vat of stuff and I'm just going to chant 'Double double, toil and trouble' and out pops my graph.There just seems to be the need for a bit of something [more] in the citation."

Isn't that the case with all serious research? A simple citation to the source does in no way reflect all the work that went into finding that source.

In this case, your discussion would explain that work and the tools you used much more clearly than a formal citation to a tool. Those tools are the same for every person, but the results are different. You can identify the tool and the website; but the result is unique to you alone and no one can go to the website for that tool and verify your results.

In EE's view, the article you are writing to defend your conclusion will or should provide the details the reader needs to understand and evaluate your results. Or, it may be necessary to write a much-more detailed "technical paper" that explains your work.

Yes, and the explanation we make would then enable anyone, who has access to the original AncestryDNA report that we cite, to replicate the processes of our "experiment" with that data and, theoretically, get the same results.

Cryptoref, I also should have asked you another question.  In your proposed first layer, you cite a title as Net1.png.  Given that italics for a title mean that we are citing a standalone publication, did you mean to use italics?

We visualize data in different ways in our reports. This visualization of the DNA matches is just another way to present the evidence from the source, similar to a table that correlates data from three census records. You cite the census resources as the source, not the table.

...census records... 

Shouldn't be typing on my phone!

As Elizabeth and Yvette have both stated, the graph is simply a visualization of sources. It does not provide any value as a source, so a citation to the image itself would not be appropriate for any textual assertion. Furthermore, its value even as a graph is limited by the sheer unwieldy volume of matches depicted. A far more informative graphic would be to localize the graph to the matches in question, so that the connections among them could actually be read. However, in this case, the graphic itself is simply a representation and would be referenced as (for example) "Figure 1. The In-Common-With Connections of Select DNA Matches," itself citing the underlying test results.

Beyond this, as a long-time student of Social Network Analysis and user of Gephi (I lectured on the subject of historical/ancestral SNA, demonstrating Gephi, at the 2016 APG Professional Management Conference), I would note that the image in question could have been produced by nearly any graphing tool. Gephi itself, on the other hand, is a powerful SNA tool that is capable of conducting complex mathematical functions related to the network structure itself. If you had been using any of these functions as part of a proof argument, those calculations could potentially have value and the tool calculating them could *possibly* be cited as a source---based on the complexity of the math involved and the fact that the software was doing them rather than you yourself. (It is, of course, possible for some of the more mathematics-oriented genealogists to read the underlying literature and perform those mathematical calculations manually.) Whether using a tool like Gephi or manually calculating the metrics oneself, however, the text would require a deep understanding of the meaning of the various metrics and their applicability to the genealogical argument being produced, in addition to the ability to express this in a comprehensible manner to an audience that may not understand any of it. In this way, SNA metrics would be similar to the use of other mathematical functions like measures of probability or, on a profoundly simpler level, calculating date of birth from a reported age. Ultimately, though, any of these mathematical calculations would be at best a form of analysis and/or correlation and insufficient to meet the Genealogical Proof Standard without meeting each criterion, esp. including conducting reasonably exhaustive research and performing other types of analysis and correlation well beyond just that represented by the mathematics.

(I apologize if I rambled a bit here, but it is a subject that I greatly enjoy discussing.)

Thanks, Yvette and Michael, for weighing in. Your perspectives greatly enrich my own humanities-based approach to the question.