Ancestry is throwing up an 1891 census entry for a Sarah W A Duncan in Cardiff, she is 7 months old (can't obviously see a birth record for a Sarah Duncan that year in Cardiff...); father William Duncan, marine engineer, b. 1853, Aberdeen; his wife, Ellen Duncan, b. 1854, Rugby...
I just checked, and the original quite clearly says Sarah, not Laura. Maybe Laura in the index is a misindexing of Sarah? However I don't immediately see a death of a matching Sarah. Also I don't seem to be able to find Laura, aged 10, in the 1901. Must be using a defective index!
Also, of course, the original does
not say that William was b 1853 or that Ellen was b 1854. The census
never gives a year of birth. In this case it says that he was 38 and she was 37, so we can deduce that he was born 1852/1853 and she was born 1853/54. This exactly matches his age in 1911, which helps to increase the confidence level in the accuracy of his age.
If Sarah/Laura was not Annabella's daughter, then she shouldn't be one of the children shown in the 1911 census as deceased, as she isn't a child of William's current marriage. On the other hand, the marriage and children details in the 1911 are on William's line of the census return, not Annabella's, so he may have misunderstood the question and included deceased children from a previous marriage.
