DNA survey - please respond

[Bob Spiers]

I have now surveyed the locations of my 4 x G Grandparents to check if this makes a deal of difference to what I reported for 3 x G. As I suspected it confirms that I was more or less on track, with 4 unknowns to take into account.

Below I shown what I reported for 3 x G with rounded percentages and the same for 4xG alongside.
(Comments follow after the statistics)

3xG (32 No.) 4xG (64 No.)
Worcestershire 13.5 (42%)……20 (31%)
Warwickshire 8 (25%)… .16(25%)
Ireland 4 (13%)……8 (13%)
Staffordshire 2 (6%)…… 4 (6%)
Cheshire 1.5 (5%)…… 3 (5%)
Herefordshire 1 (3%)…… 4 (6%)
Gloucestershire 0.5 (1.5%)… 0 (0%)
Shropshire 0.5 (1.5%).... 2 (3%)
Lancashire 0.5 (1.5%).... 3 (5%)
Wales 0.5 (1.5%).... 0 (0%)
Unknown --- ..... 4 (6%)

Apologies if figures do not align after posting

Warwickshire, Ireland, Staffordshire & Cheshire were on 'par' (double the number and the same percentage) . Herefordshire gained at the expensive of both Worcestershire Gloucestershire & Wales, as did Shropshire from Worcestershire. Lancashire gained by swings and roundabouts, but mostly from Warwickshire which would otherwise have shown an improved percentage.

As for the 4 Unknown with bare bone information, my gut feeling is 2 will probably be assigned to Worcester and 2 to Warwickshire. But all in all no great surprises. The 3 main Midland counties where I expected my ancestry to lie -Warwickshire, Worcestershire & Staffordshire down 9% from 3G (even though I expect the unknowns to eventually improve this showing) but Cheshire, Herefordshire, Shropshire & Lancashire now showing more strongly.

Sorry if it all sounds a bit like a political analysis of an Election.

 

[trebor]

I am not sure what this analysis will achieve but here are my results. I have added to the spreadsheet the details of my 3x GG.
The birth locations of these confirm my selection of the 3 main counties = Paternal side Durham & Yorks - Maternal side Staffs
However while Staffs is #2 on my list (24 pages) (Bedfordshire with no known connection and 8 pages #1) and Durham #8 (13 pages) - Yorks is a lowly #37 (36 pages) - probably due to the mix of the three Ridings as Peter has already mentioned.
With 7 of my 3x GG being totally unknown these comparisons could be quite inaccurate.

My successes with DNA matches so far are all from my Paternal side (all bar 1 Durham) with no joy at all on my Maternal side which is in line with connections made through other sites / methods.
In the 660 pages I have in total there is a very high proportion of entries with either no tree or only a handful of people in the existing ones - not sure how this affects the results.
In my county lists there is a total of 501 pages but this is not an acceptable comparison as many of the entries appear in more than one county.

I look forward to seeing any comments that can be made.

 

[Charlotte26]

I have read the initial instructions for taking part and glanced at the subsequent replies but think I must be missing something here. If I just search my matches for counties it will filter the results to produce lists of matches where that county appears in their family tree. The county in question may only appear once in that particular family tree and does not necessarily have any relationship to my direct ancestral line. Is this analysis intended to be used only on proven Ancestry DNA matches where the actual connection is known otherwise it is just indicating past population spread throughout the UK?

 

[peter]

I have read the initial instructions for taking part and glanced at the subsequent replies but think I must be missing something here. If I just search my matches for counties it will filter the results to produce lists of matches where that county appears in their family tree. The county in question may only appear once in that particular family tree and does not necessarily have any relationship to my direct ancestral line.

There will be some that match purely by chance - but because there are so many matches they won't prevent a statistical analysis from producing meaningful results.

Is this analysis intended to be used only on proven Ancestry DNA matches where the actual connection is known otherwise it is just indicating past population spread throughout the UK?

Although half of your matches might be spurious, there is no way of knowing which half. But fortunately it doesn't matter - as you can see from the spreadsheets there are big differences between the results that different members are getting depending on their ancestry.

 

[peter]

OK, here's an algorithm which not only makes logical sense but also produces much better results in most cases. What I've done in the final column is take the square root of the population of each county.

I'm attaching a revised spreadsheet with my results; you can either adapt your existing spreadsheet or copy your results into mine.

Thanks to everyone who is taking part!

 

[Bob Spiers]

Have had a quick look and adjusted and seems to have uplifted some and down played others, but the main ones are at the top. Have to dash now as about to holiday for a week, so will see how others report and reveal my own when I return.

 

[Charlotte26]

All of my English direct ancestors as far back as x5 and x6 great grandparents were born in one county although due to pre 1844 boundaries there may be some slight differences. These are only my maternal ancestors though but even so these reults are surprising.

 

[trebor]

Not sure what the new calculation does but it has certainly made what appear to be major changes to my results.
Staffs has been promoted from #2 to #1
Durham has remained at #8
Yorks has made a mighty leap from #37 to #4
Former #1 Beds (with no known connections) has been demoted to #12
I wish it would make a difference to the actual matches I have found on my Maternal side!

Sounds like Saturday afternoon listening to the football results

 

[peter]

Under the original algorithm large countries were at a significant disadvantage - the new algorithm ought to be more realistic and it looks so far as if it is.

I'd like to think that eventually we'll have something that works as well as Living DNA's ethnicity estimates do for England. And then, perhaps, we can tackle other countries....

 

[jorghes]

I've adjusted for the new algorithm and finished adding all the different results I already have (they can be accessed on the sheet tabs across the bottom of the spreadsheet).

I think while the new algorithm seems to work well for me, my mother (maternal) and both my siblings (B1 and B2), it is not as successful for my father (paternal) or my grandmother (PG).
This would most likely because my paternal grandmother (as previously mentioned) only has a quarter English ancestry - and that was vastly mobile - the rest of her ancestry is European Jewish, Scottish and Ulster Scots/Irish. My father's has this influence as well another large portion of Scottish ancestry from my paternal grandfather.

It should be noted that my mother's ancestry has a large portion of Welsh, but considering it's from the Welsh Borders, it's probably reasonable for the purposes of this algorithm.

 

[Sue345]

My revised list. I have Merchants in my Maternal ancestry who traded in London as well as their home towns and married people from all over the country who they met in London.

 

[peter]

Thanks to Sue345 and jorghes - any chance you could graduate the highlighting to indicate which of the counties account for the more or less of your ancestry (see my spreadsheet for an example).

 

[Pauline]

Here are my adjusted results, along with those my my husband(A) and my sister (J). It's worked well for my husband but not so well for me and my sister - Kent & Wiltshire are nicely at the top but Oxfordshire, Gloucestershire and Lincolnshire are too low in the list.

To be fair, I haven't yet identified any closer matches in Lincolnshire so maybe we just haven't inherited much Lincolnshire DNA.

 

[Sue345]

Version 3...

 

[peter]

it is not as successful for my father (paternal) or my grandmother (PG).
This would most likely because my paternal grandmother (as previously mentioned) only has a quarter English ancestry - and that was vastly mobile - the rest of her ancestry is European Jewish, Scottish and Ulster Scots/Irish. My father's has this influence as well another large portion of Scottish ancestry from my paternal grandfather.

I'm sure that's the reason - if only one quarter of your paternal grandmother's ancestry is presumed English the signal to noise ratio will be 4 times lower, since the algorithm is only looking at England.

Had your paternal grandmother been like my wife it might have been different. My wife is 3/4 Welsh, but all of her English ancestors came from Gloucestershire - and that's the county at the top of the list.

 

[peter]

Here are my adjusted results, along with those my my husband(A) and my sister (J). It's worked well for my husband but not so well for me and my sister - Kent & Wiltshire are nicely at the top but Oxfordshire, Gloucestershibre and Lincolnshire are too low in the list.

To be fair, I haven't yet identified any closer matches in Lincolnshire so maybe we just haven't inherited much Lincolnshire DNA.

That's always going to be a problem, unfortunately - we don't inherit DNA equally from all of our ancestors (other than our parents).

 

[jorghes]

Had your paternal grandmother been like my wife it might have been different. My wife is 3/4 Welsh, but all of her English ancestors came from Gloucestershire - and that's the county at the top of the list.

That would be absolutely brilliant if all my paternal grandfather's ancestors came from the same part of England... unfortunately a lot of them emigrated and then married, so at this point we have four (possibly more) confirmed counties - Hampshire, Yorkshire, Sussex and an ancestor who was born in London... and who knows where they came from as I have had no success figuring it out so far! (It would be brilliant if ANY of my ancestors came from a concentrated part of ANY county... and stopped moving around and making my research difficult!)

I will do my best to add gradients to the colouring, but it may take me a while to construct an appropriate view of a better set of 4th/5th great-grandparents, as that is where my research has, quite often, ground to a complete and utter halt.

 

[pjd]

Here are my adjusted results - still don't fit well but probably because the max no of recorded match pages is 40 & might well be substantially more!!

 

[peter]

Here are my adjusted results - still don't fit well but probably because the max no of recorded match pages is 40 & might well be substantially more!!

I suspect you're right - unfortunately I can't think of any way around that limitation, but maybe Ancestry will remove it at some point in the future.

 

[PhilGee]

.... - any chance you could graduate the highlighting to indicate which of the counties account for the more or less of your ancestry .....

I was about to attach my "new format" results when I saw the above. On adding "columns" for numbers of 1gt to 5gt direct ancestors for my reference (though I have only identified 30 for both 3gt and 4gt) I realised there was an anomaly (= error) for one branch where the "place of birth" was based on Parish records, supported by the Civil Registration District of Farringdon, which state "Parish of Langford" "County of Berkshire" whereas census returns correctly give "Langford, Oxfordshire".

Phil