How can I sum values at specific point

First , I have a thousand height values depending on position values. I want to order these values according to a rule. Because some height values are at the exact positon or same position. I need to sum these height values and express it with position values or plot it. For ınstance

P=[1 2 4 1 3 2 2 5]

H=[3 1 1 2 3 1 4 2]

I want these vectors

P_ordered=[1 2 3 4 5]

H_sum =[5 6 3 1 2]

Second,now, I have some other height values depends on position values named H_original and I want to do

H_original-H_sum = H_final

and plot(P,H_final,)

P=[-10 -9 -8 -7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7 8 9 10]

H_original=[10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 ]

What I want to reach

P=[-10 -9 -8 -7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7 8 9 10]

H_final=[10 10 10 10 10 10 10 10 10 10 5 4 7 9 8 10 10 10 10 10 10 ]

Plot(P,H_final)

[the cyclist]

I don't understand the rule for the second part of what you want to do.

Answer 1

[the cyclist]

First part:

If P is always some ordering of the numbers 1-N, then this simple code will work:

P_ordered = unique(P);
H_sum = accumarray(P',H')';

If not, can you please supply a more general example of how P might look?

[ak42]

I tried to explain it in the comment below

Answer 2

[Roger Stafford]

I have called the larger P array, 'P_large', to distinguish it from the original P. The following code assumes that each element of P_ordered will match one and only one element of P_large. It also assumes that P and H are the same size, and that P_large and H_original are the same size.

   P = [1 2 4 1 3 2 2 5];
   H = [3 1 1 2 3 1 4 2];
   P_large  =[-10 -9 -8 -7 -6 -5 -4 -3 -2 -1  0  1  2  3  4  5  6  7  8  9 10];
   H_original=[10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10];

   [P_ordered,~,iP] = unique(P);
   H_sum = accumarray(iP,H,size(P_ordered));
   [~,iH] = ismember(P_ordered,P_large);
   H_final = H_original-accumarray(iH,H_sum,size(P_large));

Note: In your example in obtaining H_final you appear to have misaligned H_sum with H_original by one position. This code assumes this was an error.

[ak42]

it gives size problem. And another problem(because of me), accumarray works for integer. Actually I simplified my question so it doesn't work. I wıll write it more complex. Assume I have two group data, x(y') and z(y) points, and y' is member of y or in the range of y. Now the question how can find z(y)=z(y)- x(y') at y'=y.

For example

for group of calculated data

y'_1= 0.023 is x_1=3.127

y'_2= -0.043 is x_2=2.324

y'_i= 0.112 is x_i=1.3418

and

for given data

y_1=-0.047 is z_1=2.432

y_2=0.023 is z_2= 5.231

Y_3=0.05 is z_3=4.548

y_n=0.1 is z_n=3.21

i and n are dimension of data or range. I want to find or match y' values in y and then corresponding to y'=~y I want to do z=z-x

in other words if ı get y'_2 value, it is close to y_1 and then I rounded it to y_1(because y' is not match and it is not near y_2). Later I do

z_1-x_2=2.432-2.324=0.088 =z_1 corresponding to point y_1

now for y_1=-0.047 is z_1=0.088 y_2=0.023 is z_2= 5.231

Y_3=0.05 is z_3=4.548

y_n=0.1 is z_n=3.21