Landscan help -- Landscan + panel data

2589
1
03-25-2014 07:03 PM
BenjaminThompson
New Contributor
Hi all,

I'm a research interested in changes in population across time, and specifically, migration. I've been interested in using the Landscan data to determine exactly where people are from year to year, but I'm a bit unfamiliar with the data. I was just wondering if you guys might be able to help me with some problems I've been having.

According to the Landscan FAQ, the data get more precise every year, and thus cross-year comparisons are not recommended. However, I was just wondering if there was some pattern to the improvements? I.e. could the improvements in precision be modeled, or possibly addressed using a time-fixed effects model, if I have panel data in other variables (I'd mostly be treating the Landscan counts as a dependent variable)? Does the model used to come up with the population counts change from year to year, or is it mostly the input data from the satellite which becomes finer and finer every year?

Again, I'm really new at this data, so if I've said anything that could be answered via referral, I'd appreciate anything pointing me in the right direction. It doesn't seem like there are a lot of resources to answer specific questions like I have online, but I could just be looking in the wrong place.

Thanks for your time,
Ben
0 Kudos
1 Reply
ShaunWalbridge
Esri Regular Contributor
Ben,

Those are good questions, and I think doing an accuracy assessment of the variation in the years of Landscan would be a great thing to have, and could probably stand up as its own publication. I'm not aware of any specific group which has done this, and in the past when I've worked with Landscan we've made some rather crude assumptions when looking at the data over time. I don't think this has been done before, in part because the Landscan data has a cost, and is less frequently used in analyses than its free counterparts like CIESIN's gridded population datasets.

I haven't done time-effects models, but I imagine that with some careful modeling this could be done meaningfully. You may want to look more into the existing literature published by users of Landscan to see if others have tried similar approaches. If you're only interested in particular parts of the world, you may be able to incorporate census data to predict expected differences and account for the error in Landscan. Overall, it sounds like a great project, but I'm not aware of anyone who's done this already.

cheers,
Shaun
0 Kudos