[R-sig-Geo] Error while using predict.sarlm

[Amitha Puranik]

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 using 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]]
> >
> > _______________________________________________
> > R-sig-Geo mailing list
> > R-sig-Geo using r-project.org
> > https://stat.ethz.ch/mailman/listinfo/r-sig-geo
> >
>
> --
> 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 using nhh.no
> https://orcid.org/0000-0003-2392-6140
> https://scholar.google.no/citations?user=AWeghB0AAAAJ&hl=en

	[[alternative HTML version deleted]]