Probability Assessments - using the WC Method

[Bob Spiers]

I would like to suggest a method for assessing “Probabilities” when carrying out Genealogical Research. I first came across the method in my work days as a way of assessing Health & Safety risks. It was also useful for interviews, particularly where others had to assess the same applicants, so we could compare markings.

It was introduced to me as the Winston Churchill (WC) Method as I understand he favoured it as a way of assessing wartime risks which may or may not be so. I believe it has merit and is based simply on awarding one of four marks: 2-4-6 or 8. What each mark represents is down to the user. For instance in H&S 2= low risk: 4=minimal risk: 6= moderate risk: 8=high risk.

For Genealogical Research I suggest: 2=most unlikely: 4=possible but unlikely: 6=possible and likely: 8 = possible and very likely, but you can adapt as necessary.

Sooner or later everyone comes across the build-up to what can become ‘Brick Walls’. I find a Probability Assessment using the WC Method helps me prioritise the course of future research.

There is no need to use intermediate markings, although in practice where I find something to support a previous option, or merely when comparing like-for-like marks, I add a ‘+’ where I think something deserves emphasis.

Please let me know if you think the WC Method has merit.

 

[AndyMick]

I agree this idea has considerable merit. I trialled a similar system (but with only 3 levels: 1 guess, 2 possible, 3 probable). In the end it complicated things and I didn't continue with it. If I remember rightly, The Master Genealogist programme does have a similar facility.

 

[Alexander Bisset]

Similar such systems use "traffic light" colours. Red = bad, orange = not so bad, yellow = fairly good, green = good. Or whatever wording works for you.

 

[Bob Spiers]

There is in fact a close parallel with Logic testing calling for "TRUE-FALSE-NOT PROVEN" responses. Example: “All cats are grey” (to be taken as true). Then being asked to consider: “I am grey, therefore I am a cat”? Answer= “Unproven” because he/she could equally be a grey dog (or whatever).

Probability testing is analogous with Logic testing and each of us in our own way applies logic when we have to work out if we have located the right ancestor or whatever it is we are comparing. The only problem with just pure logic is the "Unproven" handle is far too wide and you end up with half a dozen or more "Unproven" candidates or situations. I think the 2-4-6-8 method (or another of equal merit) is a judgement aid and helps concentrate the mind, and may even point the way forward.

 

[Mike]

I think the GEDCOM standard includes a Quality of Evidence tag (QUAY).

The QUAY tag's value conveys the submitter's quantitative evaluation of the credibility of a piece of information, based upon its supporting evidence. Some systems use this feature to rank multiple conflicting opinions for display of most likely information first. It is not intended to eliminate the receiver's need to evaluate the evidence for themselves.

0 =Unreliable evidence or estimated data
1 =Questionable reliability of evidence (interviews, census, oral genealogies, or potential for bias for example, an autobiography)
2 =Secondary evidence, data officially recorded sometime after event
3 =Direct and primary evidence used, or by dominance of the evidence

Not sure how well supported it is though or if it is the same thing you are talking about. I use it occasionally for "unreliable evidence" sources. i.e. a complete guess!

 

[Alexander Bisset]

Yes a QUAY tag is part of the SOUR (source_citation) record.

SOURCE_CITATION: =
[
n SOUR @<XREF:SOUR>@ /* pointer to source record */ {1:1}
+1 PAGE <WHERE_WITHIN_SOURCE> {0:1}
+1 EVEN <EVENT_TYPE_CITED_FROM> {0:1}
+2 ROLE <ROLE_IN_EVENT> {0:1}
+1 DATA {0:1}
+2 DATE <ENTRY_RECORDING_DATE> {0:1}
+2 TEXT <TEXT_FROM_SOURCE> {0:M}
+3 [ CONC | CONT ] <TEXT_FROM_SOURCE> {0:M}
+1 QUAY <CERTAINTY_ASSESSMENT> {0:1}
+1 <<MULTIMEDIA_LINK>> {0:M}
+1 <<NOTE_STRUCTURE>> {0:M}
| /* Systems not using source records */
n SOUR <SOURCE_DESCRIPTION> {1:1}
+1 [ CONC | CONT ] <SOURCE_DESCRIPTION> {0:M}
+1 TEXT <TEXT_FROM_SOURCE> {0:M}
+2 [CONC | CONT ] <TEXT_FROM_SOURCE> {0:M}
+1 <<NOTE_STRUCTURE>> {0:M}
]

Where the list that Mike provided is the possible values for QUAY. The point is that if you use these values in your source citations then assuming your Genealogy program understands them you will have a GEDCOM compliant code that will be valid in other programs.

Note that unfortunately many Genealogy programs ignore the "uncommon" parts of the GEDCOM specification and make up their own way of doing stuff. As a result many programs may not accurately record in a GEDCOM some of the more obscure types of GEDCOM. That said I've recently been made aware that some programs don't even correctly understand a GEDCOM tag of CENS (for census)!!!

 

[Bob Spiers]

As informative as both QUAY & SOUR are – and they are – it does go somewhat astray from my original intention to suggest a method of testing for probability. This occurs when seeking Ancestor ‘x’ and comparing ‘n’ findings which individually are neither ‘true’ or ‘false. (therefore unproven). A method of applying a probability rating might prove useful. Here is a pure invented example to show what I mean:


Default statement: (TRUE): Thomas Evans 1885, Miner, Rhondda,Wales

Finding:
A: ? Evans born 1888, Miner, Treorchy, South Wales= Unproven
B: Thomas Evans, 1886,?Occupation; South Wales, = Unproven
C: Thomas Evans, 1885, Miner, Nottinghamshire, England=Unproven
D:Thomas Evans, 1887, ?Occupation -Rhondda Fawr, Wales (also known father & brother are miners)=Unproven

Probability test

A: = 6+ (Treorchy is part of Rhondda, year just acceptable, need to establish first name)
B: = 4 (Need to establish occupation and Welsh birth/residence area)
C: = 2 (Unacceptable unless can establish he moved)
D: =8 (Close but need to establish also a miner)

Further research suggested by results: Concentration on D & A. If inconclusive B and/or C.

Note further research may cause Probability upgrade. For instance finding A is Thomas could upgrade his rating to 8 . Finding D to be a Miner perhaps an upgrade to 8+ In truth both are now candidates for seeking supportive proofs.