Charles Ellson wrote: > On Mon, 19 Aug 2013 15:42:44 +0100, Ian Goddard >> just that using a >> single month has a two thirds chance of being wrong even if it's what >> the official lists say. >> > Using the quarter of registration to indicate the birthdate only drops > the potential error from c.66% to c.47% (assuming for simple > convenience an even distribution across the time available for > registration). > You'll need to walk me through this one, Charles. Even distribution was one of my assumptions too. The other is that the nominal months (Mar, Jun, Sep, Dec) have 1 in 3 of the available days (2 @ 31 days plus 2 @ 30 days comes to 122 days which is as near to a third of the days in a year as you can get). Therefore 1/3 of events will fall into the nominal months of the quarters and 2/3 into the other months. Those 2/3 are in error. So how do you calculate a 47% error? -- Ian The Hotmail address is my spam-bin. Real mail address is iang at austonley org uk

On Tue, 20 Aug 2013 09:52:14 +0100, Ian Goddard <goddai01@hotmail.co.uk> wrote: >Charles Ellson wrote: >> On Mon, 19 Aug 2013 15:42:44 +0100, Ian Goddard >>> just that using a >>> single month has a two thirds chance of being wrong even if it's what >>> the official lists say. >>> >> Using the quarter of registration to indicate the birthdate only drops >> the potential error from c.66% to c.47% (assuming for simple >> convenience an even distribution across the time available for >> registration). >> > >You'll need to walk me through this one, Charles. > >Even distribution was one of my assumptions too. > >The other is that the nominal months (Mar, Jun, Sep, Dec) have 1 in 3 of >the available days (2 @ 31 days plus 2 @ 30 days comes to 122 days which >is as near to a third of the days in a year as you can get). > But the records use (or rather they used to use) quarters not months, the month indicates the "quarter ending ....."; the confusion is due to the failure of Ancestry etc. to properly identify that they are referring to quarters rather than months. It is similar to the errors encountered with ages derived from censuses which fail to allow for human beings not having their ages defined in the same way as race horses who all have their birthday on the 1st January (see also company reports, tax years etc.). >Therefore 1/3 of events will fall into the nominal months of the >quarters and 2/3 into the other months. Those 2/3 are in error. > No, the event could have occurred up to 42 days (6 weeks) before the quarter of registration thus a birth properly registered in MAR quarter could have occurred as long ago as the previous November. >So how do you calculate a 47% error? > A combination of distraction and miskeying I suspect. For MAR quarter the nominal period is c.13 weeks but the events _registered_ could have occurred any time during a c.19 week period. 6/19 = 31.6% from the previous quarter (and year). That deals with the available dates of birth rather than the actual probability of the birth being in the previous quarter which can be expected to be a lower figure if diligent registrars and parents not being slow to register are assumed.

Charles Ellson wrote: > On Tue, 20 Aug 2013 09:52:14 +0100, Ian Goddard > <goddai01@hotmail.co.uk> wrote: > >> Charles Ellson wrote: >>> On Mon, 19 Aug 2013 15:42:44 +0100, Ian Goddard >>>> just that using a >>>> single month has a two thirds chance of being wrong even if it's what >>>> the official lists say. >>>> >>> Using the quarter of registration to indicate the birthdate only drops >>> the potential error from c.66% to c.47% (assuming for simple >>> convenience an even distribution across the time available for >>> registration). >>> >> >> You'll need to walk me through this one, Charles. >> >> Even distribution was one of my assumptions too. >> >> The other is that the nominal months (Mar, Jun, Sep, Dec) have 1 in 3 of >> the available days (2 @ 31 days plus 2 @ 30 days comes to 122 days which >> is as near to a third of the days in a year as you can get). >> > But the records use (or rather they used to use) quarters not months, > the month indicates the "quarter ending ....."; the confusion is due > to the failure of Ancestry etc. to properly identify that they are > referring to quarters rather than months. It is similar to the errors > encountered with ages derived from censuses which fail to allow for > human beings not having their ages defined in the same way as race > horses who all have their birthday on the 1st January (see also > company reports, tax years etc.). > >> Therefore 1/3 of events will fall into the nominal months of the >> quarters and 2/3 into the other months. Those 2/3 are in error. >> > No, the event could have occurred up to 42 days (6 weeks) before the > quarter of registration thus a birth properly registered in MAR > quarter could have occurred as long ago as the previous November. > >> So how do you calculate a 47% error? >> > A combination of distraction and miskeying I suspect. For MAR quarter > the nominal period is c.13 weeks but the events _registered_ could > have occurred any time during a c.19 week period. 6/19 = 31.6% from > the previous quarter (and year). That deals with the available dates > of birth rather than the actual probability of the birth being in the > previous quarter which can be expected to be a lower figure if > diligent registrars and parents not being slow to register are > assumed. > Two issues here (apart from the fact that I should have said prompt registration was also an assumption). One is that, as you point out, "could have" is not the same as "invariably did". And different events are affected in different ways. As the registration was rolled into the marriage ceremony so unless the returns were not made on time there should not be a discrepancy - and one imagines that registrars in the habit of making late returns would have had words said to them. I'm not sure at what stage a death certificate was required for burial but when that became the norm it would have been difficult to avoid registering the death more than a few days after the event. That just leaves births. I had a quick check on a few births in my own family from late C19th & early C10th) where I have dates of birth readily to hand. There were two births within the last 6 weeks and both were registered in the correct period on one side. On the other side, rather more remote from the registration office, there were four in the last 6 weeks of which one was in the correct period (actually it's maybe not surprising that a child born on the 28th of December missed registration). So on a very small sample it looks as if half were registered late and that logistics could have been a factor in this; it would be interesting to see how representative this is of the wider population. Even allowing for the worst case on births the overall errors will be much less due to the other events having much less scope for errors. The other issue is that, taking the extreme case of *all* events being reported 6 weeks late, you original calculation is the correct one. To check this, consider how many results are correct in a 13 week return: it's those from the the first 7 weeks. So 7/13 of the quarter's return are correct and it follows that 6/13 are errors. -- Ian The Hotmail address is my spam-bin. Real mail address is iang at austonley org uk

On Thu, 22 Aug 2013 00:01:42 +0100, Ian Goddard <goddai01@hotmail.co.uk> wrote: >Charles Ellson wrote: >> On Tue, 20 Aug 2013 09:52:14 +0100, Ian Goddard >> <goddai01@hotmail.co.uk> wrote: >> >>> Charles Ellson wrote: >>>> On Mon, 19 Aug 2013 15:42:44 +0100, Ian Goddard >>>>> just that using a >>>>> single month has a two thirds chance of being wrong even if it's what >>>>> the official lists say. >>>>> >>>> Using the quarter of registration to indicate the birthdate only drops >>>> the potential error from c.66% to c.47% (assuming for simple >>>> convenience an even distribution across the time available for >>>> registration). >>>> >>> >>> You'll need to walk me through this one, Charles. >>> >>> Even distribution was one of my assumptions too. >>> >>> The other is that the nominal months (Mar, Jun, Sep, Dec) have 1 in 3 of >>> the available days (2 @ 31 days plus 2 @ 30 days comes to 122 days which >>> is as near to a third of the days in a year as you can get). >>> >> But the records use (or rather they used to use) quarters not months, >> the month indicates the "quarter ending ....."; the confusion is due >> to the failure of Ancestry etc. to properly identify that they are >> referring to quarters rather than months. It is similar to the errors >> encountered with ages derived from censuses which fail to allow for >> human beings not having their ages defined in the same way as race >> horses who all have their birthday on the 1st January (see also >> company reports, tax years etc.). >> >>> Therefore 1/3 of events will fall into the nominal months of the >>> quarters and 2/3 into the other months. Those 2/3 are in error. >>> >> No, the event could have occurred up to 42 days (6 weeks) before the >> quarter of registration thus a birth properly registered in MAR >> quarter could have occurred as long ago as the previous November. >> >>> So how do you calculate a 47% error? >>> >> A combination of distraction and miskeying I suspect. For MAR quarter >> the nominal period is c.13 weeks but the events _registered_ could >> have occurred any time during a c.19 week period. 6/19 = 31.6% from >> the previous quarter (and year). That deals with the available dates >> of birth rather than the actual probability of the birth being in the >> previous quarter which can be expected to be a lower figure if >> diligent registrars and parents not being slow to register are >> assumed. >> > >Two issues here (apart from the fact that I should have said prompt >registration was also an assumption). > >One is that, as you point out, "could have" is not the same as >"invariably did". And different events are affected in different ways. > As the registration was rolled into the marriage ceremony so unless >the returns were not made on time there should not be a discrepancy - >and one imagines that registrars in the habit of making late returns >would have had words said to them. I'm not sure at what stage a death >certificate was required for burial > The Births and Deaths Registration Act 1836 required that a death be registered (or a Coroner's certificate be issued) otherwise "if any dead Body shall be buried for which no such Certificate shall have been so delivered, the Person who shall bury or perform any Funeral or any religious Service for the Burial shall forthwith give Notice thereof to the Registrar ...... and who shall not within Seven Days give Notice thereof to the Registrar, shall forfeit and pay any Sum not exceeding Ten Pounds for every such Offence." s.1 Births and Deaths Registration Act 1926 specifically prohibits disposal of a body without a registrar's certificate or coroner's order. The Registration of Births and Deaths Act 1874 does not have an equivalent section. The Burial Laws Amendment Act 1880 seems to have continued the 1836 arrangements. If The Burial and Registration Acts (Doubts Removal) Act 1881 (repealed by the 1926 Act but not found via Google) did not alter this then the answer is possibly 1926. I have however got a relative who was knocked down by a lorry and taken to hospital in Finchley in the 1940s and buried the same month in Scotland but with no trace in either the deaths index or the local coroner's records; it would seem unlikely that she was rushed over the border to breath her last thus avoiding issue of an "out of England" certificate. >but when that became the norm it >would have been difficult to avoid registering the death more than a few >days after the event. > >That just leaves births. I had a quick check on a few births in my own >family from late C19th & early C10th) where I have dates of birth >readily to hand. There were two births within the last 6 weeks and both >were registered in the correct period on one side. On the other side, >rather more remote from the registration office, there were four in the >last 6 weeks of which one was in the correct period (actually it's >maybe not surprising that a child born on the 28th of December missed >registration). So on a very small sample it looks as if half were >registered late and that logistics could have been a factor in this; it >would be interesting to see how representative this is of the wider >population. > >Even allowing for the worst case on births the overall errors will be >much less due to the other events having much less scope for errors. > >The other issue is that, taking the extreme case of *all* events being >reported 6 weeks late, you original calculation is the correct one. To >check this, consider how many results are correct in a 13 week return: >it's those from the the first 7 weeks. So 7/13 of the quarter's return >are correct and it follows that 6/13 are errors.