[R-sig-Geo] problem on converting the coordinates into interval [0, 1]?

[Roger Bivand]

On Mon, 24 Sep 2007, zhijie zhang wrote:

> Dear Roger,
>  The reason for doing this is to get more general programs, which can be
> used in the other similar conditions. If we can convert all the related
> objects into internal [0,1], we can more easily use the intermediate results
> calculated from unit square,such as .bandwidth in kernels.

Exactly - with the original data, the units *are* general, because they 
are in meters. Provided the other data is also using the same metric, this
also permits a direct interpretation of bandwidth, possibly with a 
physical or biological basis. In the same sense, you can also get a 
relative bandwidth from (bw / max(diff(bbox(<obj>)))), and multiply it out 
again for a different context.

>  I think the main problem maybe the lines object, that is qiupu.shp. The
> scaling factor on the longer axis is from the guichi polygon.
> #read the lines of two rivers in the guichi
> rivers <- readShapeLines("e:/qiupu.shp")  #change the coordinates??difficult
> #plot(rivers,add=T)
> I'm not very sure if they,especially qiupu lines, can be successfullyy
> done.

Just loops of lapply(), not worse:

lns <- slot(rivers, "lines")

new_lns <- lapply(lns, function(x) {
   Lns <- slot(x, "Lines")
   new_Lns <- lapply(Lns, function(y) {
     new_crds <- mess_up(slot(y, "coords"))
     Line(new_crds)
   })
   Lines(new_Lns, ID=slot(x, "ID")
})
new_rivers <- SpatialLines(new_lns)

for your mess_up() - note, untried. Also note that the projection will not 
be defined.

Roger

> Thanks very much.
>
> On 9/24/07, Roger Bivand <Roger.Bivand at nhh.no> wrote:
>>
>> Dear Zhijie Zhang,
>>
>> On Sun, 23 Sep 2007, zhijie zhang wrote:
>>
>>> Dear Roger,
>>>  I have a problem on converting the coordinates into interval [0,1],
>> hoping
>>> that you can give me a hand.
>>
>> I think that you need to explain why. If it is just because the
>> coordinates "explode" because they are large, then simply scaling like
>> this:
>>
>> library(rgdal)
>> guichi <- readOGR(".", "guichi")
>> plot(guichi, axes=TRUE)
>> bbox(guichi)
>> proj4string(guichi)
>> t1 <- paste("+proj=tmerc +lat_0=0 +lon_0=117 +k=1.000000 +x_0=0",
>>   "+y_0=-3348000 +a=6378140 +b=6356755.288157528 +units=km")
>> t2 <- spTransform(guichi, CRS(t1))
>> plot(t2, axes=TRUE)
>> bbox(t2)
>> #          min      max
>> # r1 10.148541 79.88203
>> # r2  0.834946 61.90978
>>
>> gets down to 70 by 80, which is not such a numerical risk (if need be, use
>> +units=kmi to get nautical miles, which reduces the numbers even more). It
>> has the key advantage that you can always get back to the true coordinates
>> by transforming back.
>>
>> In the internal code, you will see that many packages guard by using [-1,
>> +1] on both dimensions. In your case and if you know that you need this,
>> you would have to work out the scaling factor on the longer axis, so that
>> points on the shorter axis would be scaled correctly. But if you want to
>> keep the dimensions' aspect, you could just shift the origin and the units
>> as I show above.
>>
>> Hope this helps,
>>
>> Roger
>>
>>>  My data consist of points(cases), lines(rivers) and polygon (guichi
>> city),
>>> and i want to change their coordinates into interval [0,1].  I have put
>> the
>>> data in the attachment, so that you can use it.
>>> The following is the programs to read the data and one possible method
>> for
>>> conversion:
>>>
>>> library(sp)
>>> library(foreign)
>>> library(mgcv)
>>> library(maptools)
>>>
>>> #read the polygon containing the studied points
>>> guichi <- readShapePoly("e:/guichi.shp")  #boundary polygons containing
>> the
>>> points
>>> point_poly <-
>>>
>> getPolygonCoordsSlot(getPolygonsPolygonsSlot(getSpPpolygonsSlot(guichi)[[1]])[[1]])
>>> #get the coordinates of guichi
>>> #plot(point_poly,xlab="x", ylab="y",type="l")
>>>
>>> #read the lines of two rivers in the guichi
>>> rivers <- readShapeLines("e:/qiupu.shp")  #change the
>> coordinates??difficult
>>> #plot(rivers,add=T)
>>>
>>> #read the points of cases and controls
>>> case_control <- read.csv("e:/casecontrol.csv",sep=",", header=TRUE)
>>> #plot(case_control$x,case_control$y)
>>>
>>> *#Plot the whole figure*
>>> plot(point_poly,xlab="x", ylab="y",type="l")
>>> plot(rivers,add=T)
>>> points(case_control$x,case_control$y,add=T)
>>>
>>>
>> ###################################################################################
>>> #one of the possible methods to convert the x/y coordinates into
>> interval
>>> [0,1]
>>> *# But it seems there is a problem: it convert the x/y coordinates into
>>> interval [0,1], respectively.
>>> # In my opinion, they should be expand or  shrink according to the same
>>> minimum/maximum value*.
>>> st <- function(x)(x-min(x))/(max(x)-min(x))
>>> case_control[,c(8,9)] <- data.frame(lapply(case_control[,c(6,7)],st))
>>>
>> ###################################################################################
>>> *Q1. The main problem is on the rivers, which has multiple lines. There
>> are
>>> difficulties for conversion.
>>> Q2. when i convert the x/y coordinates, i'm not very sure whether i
>> should
>>> use the same scale of conversion or not. *
>>> Could you please help us to solve it?
>>> Any suggestions/help are greatly appreciated.
>>>
>>>
>>>
>>
>> --
>> Roger Bivand
>> Economic Geography Section, Department of Economics, Norwegian School of
>> Economics and Business Administration, Helleveien 30, N-5045 Bergen,
>> Norway. voice: +47 55 95 93 55; fax +47 55 95 95 43
>> e-mail: Roger.Bivand at nhh.no
>>
>
>
>
>

-- 
Roger Bivand
Economic Geography Section, Department of Economics, Norwegian School of
Economics and Business Administration, Helleveien 30, N-5045 Bergen,
Norway. voice: +47 55 95 93 55; fax +47 55 95 95 43
e-mail: Roger.Bivand at nhh.no