3

Is it possible to do stochastic population forecasting with the R package Demography when the population size is between 10k and 20k? I have data from year 1985 to 2012 for population, deaths and births. What do you do when almost half of the death rates are 0? From ages 1 to 30 nearly no one dies in a so small population. Is there a method for imputing some plausible values for the zero's in the death and birth rates?

Is it possible to, for example, multiply all population numbers with 1000 and impute some plausible values for rates (maybe borrow these from a similar population), and then divide all the simulated numbers in output with 1000?

I have tried to do some forecasting with the R package demography, but I get so many warnings and weird results, I can't get it right. Is there a population size limit for using the demography package?

Do anyone have any suggestions for how to solve the problem of stochastic forecasting with small populations?

gung - Reinstate Monica
  • 132,789
  • 81
  • 357
  • 650
Hdmas
  • 31
  • 1
  • 3

1 Answers1

1

For the zero death rates, the package won't work as it has to take log of the death rates and obviously you can't take log of zero.

You can set the values to missing but it is incorrect as you do have the values!

You can also try to put a very very very small number, but here again you are going to 'fake' the data.

What you can do is to try to group the ages in such a way that you never have zero values.

Akshav
  • 11
  • 3