Oman 2015 data

[dkobak]

Leaving it here so that we don't forget: 2015 data for Oman looks like a low outlier, suggesting maybe incomplete registration in 2015. The subsequent years (2016-2019) do not show any trend. We may consider to exclude 2015 values; this will noticeably lower our baseline projection for 2020.

[akarlinsky]

I contacted Oman's NSO and will update if and when they answer me. For now I'm keeping this as-is.

[st2048]

I have checked Oman's 2015 report on births and deaths and I have noticed the sum of monthly values is significantly lower than the yearly figure.

I can see the monthly values on Tables 34 (Omanis) and 35 (Expatriates) - the sum of the values for Omanis and Expatriates is equal to the values in your dataset. In these tables, the sum of deaths for all months is given as 5942 for Omanis and 1401 for Expatriates, leading to the grand total of 7343.

However, that same report mentions the yearly figures for registered deaths in 2015 in Table 21, which are 6832 for Omanis and 1335 for Expatriates, leading to a grand total of 8167.

This discrepancy does not seem to happen in other years. For example, this is the link to the 2018 report, where the monthly data from Tables 15-16 is perfectly consistent with the yearly totals of Table 1. The data on pages 30-31 comparing 2018 data to earlier years also seems to suggest they are using the value of 8167 for 2015 - in turn, 2016-2018 values are consistent with your dataset.

I can see in the description that you scaled up the 2019 monthly data to match the yearly data. Have you considered the possibility of doing the same for 2015?

[dkobak]

That's a very perceptive observation.

By the way, Ariel, I don't actually like this adjustment procedure very much:

Note: 2019 monthly death counts was smaller at 1020 deaths than total yearly mortality count (Shown in the January 2021 Monthly Bulletin). In order to account for this, all 2019 monthly data was increased by the mean monthly difference of 85. Final official monthly data for 2019 should be available by November 2021.

I think we should not add 85 to each month, but scale all 2019 monthly values by the ratio yearly_value/sum_monthly_values. This will give non-integer values, which in this case is good (similar to what we have for Sweden etc.)

[akarlinsky]

@st2048 I just did that myself before saw you posted this to make sure, I agree 100%.
My query to Oman's NSO is thus requesting monthly counts for 2015 to 2019 to see if maybe the 2015 yearly report is just much less updated.
Regardless, there's still a large jump between 2015 to 2016, from 8167 to 8829 so trend will still prob. be "too-high".

I will update when I get some response.

[dkobak]

I assume NSO never replied. Anyway - is there any reason not to scale 2015 monthly values to make them add up to the 2015 annual value, similar to how we do it for 2019?

[dkobak]

Getting back to this issue:

I assume NSO never replied. Anyway - is there any reason not to scale 2015 monthly values to make them add up to the 2015 annual value, similar to how we do it for 2019?