search

ITHIM / ITHIM-R

Development of the ITHIM-R, also known as ITHIM version 3.0. Started in January 2018.
https://ithim.github.io/ITHIM-R/
GNU General Public License v3.0
17 stars 11 forks source link

Population numbers-MSLT #33

Open BelenZapata85 opened 6 years ago

BelenZapata85 commented 6 years ago

For the PMSLT we need population numbers per one year age group and sex for a given year from age 0 to 100. That is to say, we need the distribution of the total population in one year age group by sex. @AnnaGoodman1 will such data be available from the synthetic population?

Thanks, Belen

AnnaGoodman1 commented 6 years ago

not in the version that I've created for metahit, I'm afraid, it only has people aged 16+, and only has them in five-year age categories. possibly you could assign a one-year age at random within the five-year age categories, but if you needed people age 0-15 you would have to recreate this population somewhat from scratch. (I could share the stata do-files I used to make it, although they are a bit complicated and I'm about to go maternity leave so wouldn't be able to help, sorry)

  Male female
16 to 19 1,501,912 1,409,630
20 to 24 1,679,888 1,693,330
25 to 29 1,619,781 1,697,836
30 to 34 1,955,658 1,886,827
35 to 39 1,699,478 1,733,095
40 to 44 1,992,422 2,019,258
45 to 49 1,916,881 1,953,731
50 to 54 1,622,863 1,688,346
55 to 59 1,444,598 1,494,450
60 to 64 1,659,153 1,659,954
65 to 69 1,202,281 1,306,021
70 to 74 983,637 1,060,344
75 to 79 663,644 1,013,242
80 to 84 496,591 763,873
85 to 89 334,510 442,230
90 to 94 127,295 178,555
95+ 36,881 51,425
BelenZapata85 commented 6 years ago

Thanks Anna, we only need adults. I will start by trying to work from the age groups, which I believe will be possible, I need to check road trauma estimates with @robj411 . Rob, is it possible to have matching estimates of numbers from your model to the age groups? thanks, Belen

AnnaGoodman1 commented 6 years ago

Hi Belen, OK great. You could use this data https://www.nomisweb.co.uk/census/2011/dc1117ew if you wanted to model more exactly the distribution of single-year ages within five-year age bands (e.g. there are probably slightly more 70 year olds than 74 year olds, because of mortality). But that is quite possibly overkill... Anna

gotom22 commented 6 years ago

@BelenZapata85

my sense is that you should develop this based on the likely available data - wich in your case, it seems realistic that users will have 1-year data for age and deaths. So I think there will be a by-pass to the aggregation steps for the synthetic popupulation, rather than a (random) re-assigning of 1-year ages.

Question for you though: what other data do you use (that is specific to the PMSLT approach) and at what level of aggregation?

@JDWoodcock

BelenZapata85 commented 6 years ago

Thanks @gotom22 (@JDWoodcock) , yes, seems realistic to think that users will have one-year population and mortality data. If not, for population data, we can work with 5-year age groups and mortality could be interpolated (interpolation method is not in the code yet).

I am working on defining data inputs for the PMSLT. For some of the data there is a two step process from GBD data to data that is suitable for the model derived with Dismod II (but other methods to be discussed). I am working on Figure 3 in here https://github.com/ITHIM/ITHIM-R/blob/master/MSLT/MSLTdoc.Rmd. The issue is that for most cities we only have country level data, so scaling options from country level to city level will be discussed with @nmaizlish

gotom22 commented 6 years ago

sounds good!

In terms of structure/terminology:

...ideally this aligns with how you code and document...