quick merge question

[Mike Tanner]

Hello!

I am trying to merge 2 data sets, one for rainfall and one for temperature. Both have information of Longitude, Latitude, Date and rainfall (obs per grid) and obviously one temp and one rainfall. there is no unique identifier, what is a goodw ay to merge them? I tried joinby but given the size of the data-sets it crashes my pc (and I am restricted to using stata).

the idea is then after having them merged, using geodist againts a database with longitude and latitude data on districts.. to drop any grid more than 100k away from the district points..

Much help appreciated

[Clyde Schechter]

Have you tried -merge 1:1 longitude latitude date- for this? If the two files match up on these three variables, that is the correct approach. If they don't, then I don't know what to tell you.

[Mike Tanner]

Thank you Clyde,

What actually worked is the command cross

cross using "district_crosswalksmall.dta"
geodist centroid_latitude centroid_longitude latitude longitude, gen (d) sphere
drop if d > 100

and now i have the data base I needed - if you dont minda follow up>

i am trying to get the weighted daily average of rainfall with an end result being a yearly panel for every district. I created the weight (weight = inverse of distance squared) - and I am to use collapse -

collapse (mean) temperature [xweight= weight], by (district id, year)

i am not sure of what type of weight I should use (thats why I typed xweight), and specifically, would this give me a weighted daily average temperaturefor a district level daily dataset from 2009 to 2013

[Mike Tanner]

I experimented a little and this gave me yearly averages,

collapse (mean) temperature [xweight= weight], by (district id, year)

so, I added the date variable and used aweights

collapse (firstnm) statename districtname (mean) temperature [aweight= weight], by (district id, edate, year)