Error while using predict.sarlm

str(attr(weightmatrix, "region.id"))

str(row.names(missed_data))

all(row.names(missed_data) %in% attr(weightmatrix, "region.id"))

Dear Prof. Roger Bivand,

Thanks a lot for providing a clarification for my query.
I used the following code and found out that the region.id in listw
object and row.names of the data do not match.

str(attr(weightmatrix, "region.id"))

 chr [1:182] "1" "2" "3" "4" "5" "6" "7" "8" "9" "10" "11" "12" "13" "14"
"15" "16" "17" ...

str(row.names(missed_data))

 chr [1:182] "142" "108" "149" "76" "8" "71" "45" "75" "173" "119" "22"
"32" "156" "221" ...

all(row.names(missed_data) %in% attr(weightmatrix, "region.id"))

[1] FALSE

 How can I change the row.names of my data 'missed_data' and align to the
listw object?


       Thanks & regards,

*Amitha Puranik*















On Sat, May 25, 2019 at 9:49 PM Roger Bivand <Roger.Bivand at nhh.no> wrote:

On Fri, 24 May 2019, Amitha Puranik wrote:

I am facing an error while using predict.sarlm to make predictions for

spatial

lag model generated using lagsarlm. I used the following code:

predicted = predict(fit.lag, listw=weightmatrix, newdata=missed_data,
pred.type="TS", zero.policy = T)

For the argument newdata, I have passed the same data missed_data which

I

used to fit the spatial lag model.

When I run the above code, I get the following error message: ?Error in
predict.sarlm(fit.lag, listw = weightmatrix, newdata = missed_data,  :
mismatch between newdata and spatial weights. newdata should have

region.id

as row.names?

The predict method has to identify the weights applying to the newdata.
So
it uses the region.id attribute of the neighbour object, and the
row.names
of the newdata object. If they do not match, it error-exits. If shp below
was read in the typical way, the default region.id may be the FID of the
input file (0, ..., (n-1)), but the default row.names of newdata may be
1,
..., n.

For example:

library(sf)

Linking to GEOS 3.7.2, GDAL 3.0.0, PROJ 6.1.0

boston_506 <- st_read(system.file(

+                                   "shapes/boston_tracts.shp",
+                                   package="spData")[1])
Reading layer `boston_tracts' from data source
`/home/rsb/lib/r_libs/spData/shapes/boston_tracts.shp' using driver `ESRI
Shapefile'
Simple feature collection with 506 features and 36 fields
geometry type:  POLYGON
dimension:      XY
bbox:           xmin: -71.52311 ymin: 42.00305 xmax: -70.63823 ymax:
42.67307
epsg (SRID):    4267
proj4string:    +proj=longlat +datum=NAD27 +no_defs

nb_q <- spdep::poly2nb(boston_506)
lw_q <- spdep::nb2listw(nb_q, style="W")
boston_489 <- boston_506[!is.na(boston_506$median),]
nb_q_489 <- spdep::poly2nb(boston_489)
lw_q_489 <- spdep::nb2listw(nb_q_489, style="W", zero.policy=TRUE)
form <- formula(log(median) ~ CRIM + ZN + INDUS + CHAS +

+                 I((NOX*10)^2) + I(RM^2) + AGE + log(DIS) +
+                 log(RAD) + TAX + PTRATIO + I(BB/100) +
+                 log(I(LSTAT/100)))

suppressPackageStartupMessages(library(spatialreg))

eigs_489 <- eigenw(lw_q_489)

SLM_489 <- lagsarlm(form, data=boston_489,

+           listw=lw_q_489, zero.policy=TRUE,
+           control=list(pre_eig=eigs_489))

nd <- boston_506[is.na(boston_506$median),]
t0 <- exp(predict(SLM_489, newdata=nd, listw=lw_q,

+                   pred.type="TS", zero.policy=TRUE))

str(attr(lw_q, "region.id"))

  chr [1:506] "1" "2" "3" "4" "5" "6" "7" "8" "9" "10" "11" "12" "13"
"14"
"15" "16" ...

str(row.names(nd))

  chr [1:17] "13" "14" "15" "17" "43" "50" "312" "313" "314" "317" "337"
"346" "355" ...

all(row.names(nd) %in% attr(lw_q, "region.id"))

[1] TRUE
# introduce a wrong row.name

row.names(nd)[1] <- "0"
all(row.names(nd) %in% attr(lw_q, "region.id"))

[1] FALSE

t0 <- exp(predict(SLM_489, newdata=nd, listw=lw_q,

+                   pred.type="TS", zero.policy=TRUE))
Error in predict.sarlm(SLM_489, newdata = nd, listw = lw_q,
   pred.type = "TS",  :
   mismatch between newdata and spatial weights. newdata should have
   region.id as row.names

In this case, the row.names of the input object to spdep::poly2nb() and
the region.id matched, as the newdata were subsetted from the same
object.
We don't know the values for your data, but you should be able to check
them. It is important that they align the data with the weights correctly
for obvious reasons.

Hope this helps,

Roger

I have obtained the weight matrix from the function below

weightMat <- function(shp){

 dnb <- knearneigh(coordinates(shp), k=4)

 dnb <- knn2nb(dnb) #create nb

 lw <- nb2listw(dnb, style="W",zero.policy=TRUE) #create lw

 return(lw)

}

To cross check and make sure there are no discrepancies, I have run the
following lines

length(weightmatrix$weights)

nrow(missed_data)

nrow(coordinates(shape))

For all the codes above, the result is 182, which is the sample size of
data.

Can anyone offer me some guidance in solving this problem? Thanks for

your

help.


      Thanks & regards,

*Amitha Puranik*

Assistant Professor,

Department of Statistics, PSPH

Phone:0820-2922407
Address:Department of Statistics,

Health Sciences Library, Level 6,

Manipal Academy of Higher Education,Manipal,Karnataka,India

An Institute of Eminence (Status Accorded by MHRD)

      [[alternative HTML version deleted]]

--
Roger Bivand
Department of Economics, Norwegian School of Economics,
Helleveien 30, N-5045 Bergen, Norway.
voice: +47 55 95 93 55; e-mail: Roger.Bivand at nhh.no
https://orcid.org/0000-0003-2392-6140
https://scholar.google.no/citations?user=AWeghB0AAAAJ&hl=en