• Only registered members can see all the forums - if you've received an invitation to join (it'll be on your My Summary page) please register NOW!

  • If you're looking for the LostCousins site please click the logo in the top left corner - these forums are for existing LostCousins members only.
  • This is the LostCousins Forum. If you were looking for the LostCousins website simply click the logo at the top left.
  • It's easier than ever before to check your entries from the 1881 Census - more details here

Ancestor Number Predictor

AndyMick

AndyMick

LostCousins Star
Back in July, before the computer gremlins attacked, I created a rough computer model of how many ancestors you might expect to find given what you know about them already. This was prompted by some discussion on the forum of how many ancestors we had located. Tim's recent comment persuaded me to make it public in its current state even though the gremlins might not let me handle the feedback - we'll see!

The model is not very good, but gives an idea of where research needs to be concentrated. For example, I have considerably fewer folks in 1911 than the model suggests I should have.

I wrote it originally in VBA (in Excel) but I converted it to Javascript to run self contained in a web page - and consequently people with the requisite computer skills can look at it and see how rotten my code is - I mean can see how it could be improved:)

Talking of Alexander reminds me, the model is for England and Wales, although I don't think it would take much to adapt for the US - sorry Scots and Canadians, but you only have the one census on the system, so it's not my fault!

So, this is the link you've all been waiting for! (Lights blue touch paper and stands back!)
 
I've been having a play with the numbers - I got the best match of 1841/1881 numbers with a start date of 1770 and 3 children. But that made my 1911 even worse :(
 
Can you clarify what you mean by made 1911 even worse. Do you mean it suggested you had loads of people you should have but dont?
 
The problem I see with the model is in the go back to come forward aspect. When you are "matching" your 1841 & 1881 stats you aren't really comparing apples with apples, more like comparing apples with washing machines.

The problem is that you really don't know and there are no constraints in the model to say whether your 1841 and 1881 stats are actually right. ie: there is nothing I can see in the model that actually uses the 1841 & 1881 stats as an input to work out how many ancestors you should have. So you can't then match your 1841 & 1881 stats to see what 1911 should be as you have no idea if 1841 and 1881 are right and small variants in the numbers say add an extra 1841 family of 6 and suddenly the whole numbering is out.

It's interesting approach but I fear there are too many variables. What should be possible on the other hand is to say you have X people alive on the 1841 census with an average age of A so you'd expect Y in 1881 and Z in 1911 given the numbers of children and life expectancy.

That way you fix one point in the jigsaw. A point that can be a known quantity. Then you can extrapolate from that known point.
 
Predicting is all fine but will never work because of the variables. Assuming 2 every generation is ok, but if the 2nd generation in only has 1 then the calculation is out all the way through.

To me it makes more sense to compare your figures on LC with each other

1841 251
1881 557
1911 769

1881 is about 2 times 1841
1911 is about 3 times 1841
 
However it does have merit in dealing with expected values. Yes it won't get it 100% right not least of which is that if you find a sibling to a 3g or 4g grandfather/mother to take Andy's example the all of a sudden you have potentially doubled all the numbers for that branch.
 
You must log in or register to reply here.
Back
Top