Fluctuations in and possible unreliability of death statistics for small areas, for small sub-groups of the population, or for short periods
Fluctuations in and possible unreliability of death statistics for small areas, for small sub-groups of the population, or for short periods
Vital Events statistics are produced from complete counts of all the events which were registered, and so are not subject to some of the kinds of errors that may affect the results of sample surveys (such as non-response bias).
However, it is important to note that the Vital Events figures for, say, a small area or a small sub-group of the population may be subject to large percentage fluctuations from (e.g.) year to year, due simply to the inevitable variability of natural events. Similarly, even the total number of deaths registered across Scotland as a whole may be subject to large percentage fluctuations if it relates to a short period (such as a week), due to both natural variation and factors (such as public holidays) which affect the arrangements for registering deaths.
In addition, in the case of figures for areas within Scotland, it should be remembered that, the smaller the area is, the more (in percentage terms) its figures may influenced by the way in which the National Records of Scotland (NRS) allocates deaths to areas, based on the information that is collected by the registration process. Further information can be found in Geographical basis of Vital Events statistics (PDF 36 Kb).
It follows that the figure for a small area, or a small sub-group, for any given year may provide an unreliable indicator of the usual annual number of events. This point is illustrated in:
Fluctuations in the death statistics for datazones (PDF 32 Kb); and
An explanation of what datazones are can be found in the 'Datazone codes' section of Geographical basis of Vital Events statistics (PDF 36 Kb).
Similarly, there can be large percentage week-to-week fluctuations in the numbers of deaths registered for Scotland as a whole, so the change over a few weeks may be an unreliable indicator of any trend in the number of events. This point is illustrated in Fluctuations in numbers of deaths by week (PDF 25 Kb).
Week-by-week figures for parts of Scotland (such as local authorities) will be subject to potentially larger percentage fluctuations, because the numbers involved are smaller. Details of this can be found in Fluctuations in and possible unreliability of death statistcs for individual local authority areas for individual weeks (PDF 25 Kb).
In addition, figures for, say, local authorities will be influenced by the way that NRS allocates deaths to areas. More information about this can be found in the 'Information available about where the person lives or lived' section of Geographical basis of Vital Events statistics (PDF 36 kb).
There can also be large percentage year-to-year fluctuations in the numbers of deaths from certain causes, so the figure for any given year (or the change between two years) may be an unreliable indictor of the level and any trend in deaths from a particular cause. This point is illustrated in Fluctuations in the statistics of deaths from particular causes (PDF 25 Kb).
Age-standardised death rates for areas within Scotland may also be subject to large percentage year-to-year fluctuations, particularly in the case of figures which relate only to certain causes of death and/or to people who were aged under 75, because their underlying numbers of deaths may be relatively small. Again, the figure for any given year (or the change between two years) may be an unreliable indicator of the level or any trend. Examples of this are given in Fluctuations in age-standardised death rates (PDF 25 Kb).
In many cases, the number of deaths fluctuates randomly from period to period, and can be represented as the outcome of a 'Poisson process'. Examples of this are given on some of the other pages. For ease of reference, 'Fluctuations in the numbers of deaths may be represented as the outcome of a Poisson processs' (PDF 25 Kb) document indicates how a likely range of random period-to-period variation can be calculated in such cases.
PDF files require Acrobat Reader. Download Acrobat Reader free 