employment: employment
In DanielBonnery/dataASPEP: X

1	data(employment)

employment

id: Id
state: State
type: State ("0"), County ("1"), Township ("2"), Municipality ("3"), Special district ("4"), School district ("5")
samplingType: Type with levels "2" and "3" collapsed: level "3" corresponds to townships and Municipalities
itemcode: (Item code. "000": summary, 001= Airport 007= Space Research and Technology (Federal), 005= Correction, 006= National Defense and International Relations(Federal), 044= Streets and Highways, 050= Housing and Community Development (Local), 052= Local Libraries, 059= Natural Resources, 061= Parks and Recreation, 012= Elementary and Secondary - Instruction, 112= Elementary and Secondary - Other Total, 014= Postal Service (Federal), 016= Higher Education - Other, 018= Higher Education - Instructional, 062= Police Protection - Officers, 162= Police-Other, 129= Welfare, 080= Sewerage, 071= Other Education (State) 081 Solid 072, Social Insurance Administration (State), 073= Financial Administration, 074= Firefighters, 124= Fire - Other, 081= Waste Management, 087= Water Transport and Terminals, 089= Other and Unallocable, 090= Liquor Stores (State), 091= Water Supply, 075= Judicial and Legal, 079= Other Government Administration, 032= Health, 092= Electric Power, 093= Gas Supply, 094= Transit)
ftemp12: ftemp12
ftpay12: ftpay12
ptemp12: ptemp12
ptpay12: ptpay12
pthours12: pthours12
ftemp12: ftemp12
ftpay12: ftpay12
ptemp12: ptemp12
ptpay12: ptpay12
pthours12: pthours12
insample: inSample
weight: finalWeight

require("ipfp")
attach(employment)
#2. create small area by state and item code
states<-data.frame(state=levels(state)[sort(unique(state))])
itemcodes<-data.frame(itemcode=levels(itemcode)[sort(unique(itemcode))])
count07<-sqldf("select * from (
               select distinct a.state, a.itemcode, sum(a.ftemp07+a.ptemp07) as margin from employment a group by a.state,a.itemcode
               union all
               select * from (select distinct  b.state, c.itemcode,0 as margin from states b,itemcodes c 
               except select distinct d.state, d.itemcode, 0 as margin from employment d))
               order by state, itemcode
               ")

count12<-sqldf("select * from (
               select distinct a.state, a.itemcode, sum(a.ftemp12+a.ptemp12) as margin from employment a group by a.state,a.itemcode
               union all
               select * from (select distinct  b.state, c.itemcode,0 as margin from states b,itemcodes c 
               except select distinct d.state, d.itemcode, 0 as margin from employment d))
               order by state, itemcode
               ")


#3. create the vector of constraints
Sample<-employment[insample,]

direct12<-sqldf("select * from (
                select distinct a.state, a.itemcode, sum(a.weight*( a.ftemp12+a.ptemp12)) as margin from Sample a group by a.state, a.itemcode
                union all
                select * from (select distinct  b.state, c.itemcode,0 as margin from states b,itemcodes c 
                except select distinct d.state, d.itemcode, 0 as margin from Sample d))
                order by state, itemcode
                ")
marginitem<-sqldf("select * from (
                  select distinct a.itemcode as code, sum(a.weight*(a.ftemp12+a.ptemp12)) as margin from Sample a group by itemcode
                  union all
                  select * from (select distinct b.itemcode as code, 0 as margin from itemcodes b 
                  except select c.itemcode as code, 0 as margin from Sample c))
                  group by code order by code
                  ")
marginstate<-sqldf("select distinct state as code,   sum(weight*(ftemp12+ptemp12)) as margin from Sample group by state order by state")
margin<-sqldf("select distinct 'intercept' as code, sum(weight*(ftemp12+ptemp12)) as margin from Sample")
y<-rbind(margin,marginstate[-1,],marginitem[-1,])[,2]
A<-t(model.matrix(~state+itemcode,data=count07))

x0=count07$margin
spree<-ipfp(y=y, A=A, x0=x0, tol = .Machine$double.eps, maxit = 1000,verbose = FALSE, full = FALSE)
image(-A)
length(spree)

REspree <-(spree-count12$margin)/count12$margin
REdirect<-(direct12$margin-count12$margin)/count12$margin
hist(REspree)
hist(REdirect)
boxplot(cbind(Direct=REdirect, Spree=REspree))
plot(REspree,REdirect)
plot(spree,x0,pch=19,cex=.5,col="blue")
abline(0,1,lwd=2)