Love your blog. Your analysis, shinys, datavis and writing are always great, i read each post religiously. I've been modelling demography for around 18 months so I've enjoyed your recent ventures into this space.
I've got a topic that I'd love to see a blog post on, a topic that i don't think has ever been blogged on (at least not that i can find). Most cohort component demographic modelling exists in an unconstrained environment, i.e. without 'carrying capacity' constraints, such as available land for housing.
So the potential blog topic has two elements:
a) How would you go about the spatio-temporal modelling of carry capacity constraints using census data (such that you can forecast sub regions)? I mention census data, but i'd be interested in how else you model this too, perhaps aerial imagery?
b) How would you go about incorporating the carrying capacity modelling into the cohort component model? How would you back-propagate the changes due to carrying capacity such that the cohort component model accounting still adds up?
I've recently done the above modelling. But you are in a different league to me, so i'd be fascinated to hear how you might go about it. For what it is worth I used a spatio-temporal GAM on population densities by small statistical regions to estimate regions of change. I transformed this into a time series index for each region - the index only changed the cohort component model in regions that have been growing very rapidly for a long period of time. Where the index changed the cohort component model i used net migration to soak up these changes such that the cohort component model accounting added up.
Sorry, that 'issue' got a little long. Eager to hear your thoughts.
Hi Peter,
Love your blog. Your analysis, shinys, datavis and writing are always great, i read each post religiously. I've been modelling demography for around 18 months so I've enjoyed your recent ventures into this space.
I've got a topic that I'd love to see a blog post on, a topic that i don't think has ever been blogged on (at least not that i can find). Most cohort component demographic modelling exists in an unconstrained environment, i.e. without 'carrying capacity' constraints, such as available land for housing.
So the potential blog topic has two elements: a) How would you go about the spatio-temporal modelling of carry capacity constraints using census data (such that you can forecast sub regions)? I mention census data, but i'd be interested in how else you model this too, perhaps aerial imagery? b) How would you go about incorporating the carrying capacity modelling into the cohort component model? How would you back-propagate the changes due to carrying capacity such that the cohort component model accounting still adds up?
I've recently done the above modelling. But you are in a different league to me, so i'd be fascinated to hear how you might go about it. For what it is worth I used a spatio-temporal GAM on population densities by small statistical regions to estimate regions of change. I transformed this into a time series index for each region - the index only changed the cohort component model in regions that have been growing very rapidly for a long period of time. Where the index changed the cohort component model i used net migration to soak up these changes such that the cohort component model accounting added up.
Sorry, that 'issue' got a little long. Eager to hear your thoughts.
Alex