Interpolation for x,y,z values

I am doing interpolation of z using x & y values ..But I am getting NAN values using interpolation how can I removed this NAN and interpolate it to gives actual Value.. I am herewith attaching my Excel file, code and error

clear all

close all

clc

a=xlsread('Book2.xlsx')

x = a(:,1);

y =a(:,2);

z = a(:,3);

xq=[800 1000 1250 1500 1750 2100 2500 3000 3500 4000 4500 5000 5500 6000 6500];

yq=(100:100:2700)'

vq = griddata(x,y,z,xq,yq)

mesh(xq,yq,vq)

hold on

plot3(x,y,z,'o')

Answer 1

[Star Strider]

The data in the file do not appear to resemble the data in the image.

For those, the fillmissing function (R2016b and later releases) is likely the best option, however it will be necessary for you to experiment to get the desired result—

a = readtable('https://www.mathworks.com/matlabcentral/answers/uploaded_files/1081705/Book2.xlsx', 'VariableNamingRule','preserve')

a = 37×3 table
      N        PCOL     TAPE (iso)
    ______    ______    __________

    799.94       660      30.28   
    800.01     359.9      37.21   
    800.08    1011.3      28.63   
      1250       660      48.64   
      1250    1552.3      28.76   
      1250       360      52.57   
    1250.2      1020      28.19   
    1749.9     359.9       47.8   
      1750    1020.3      26.61   
      1750      2234      30.51   
    1750.1       660      48.77   
      2499     839.9      28.27   
    2499.7    2317.7      31.83   
    2500.7       360      35.03   
    2500.9    2312.7      31.92   
    2501.1    1020.2      26.76   

a(:,1:3) = fillmissing(a(:,1:3), 'makima', 'EndValues','extrap')        % Experiment With The 'method' (Here: 'makima'), The 'EndValues' extrapolation use the same 'method'
x = a{:,1};
y = a{:,2};
z = a{:,3};
L = size(a,1);
xq = linspace(min(x), max(x), L);
yq = linspace(min(y), max(y), L);
[X,Y] = ndgrid(xq, yq);
Z = griddata(x, y, z, X, Y);

figure
surfc(X, Y, Z)
grid on

.

[Prasad Joshi]

Hi it showed following error .Thank for the answer

[Star Strider]

There are version differences, and your release/version was not stated. This works in R2022a.

Check the 'method' options for your version/release, and choose the one that gives the best result.

[Prasad Joshi]

Its 2016b

[Star Strider]

The available Archived MathWorks Documentation does not have information for R2016b, so assuming that the behaviour did not change betweeen R2016b and R2017a, the ‘pchip’ method might be closest to 'makima'. It would be best for you to review the documentation on fillmissing (here for R2017a) for R2016b and choose the 'method' that gives you the result you want.

(MathWorks — I realise that disk space is finite, however it would really help to have access to the documentaztion on more of the previous versions!)

.

[Prasad Joshi]

Thanks

[Star Strider]

As always, my pleasure!

Answer 2

[Mathias Smeets]

You are getting NaN points because some query points (for example your lowest y-values) are outside your actual data. It is not possible to extrapolate with the griddata function. Look this link for more information.

[Prasad Joshi]

Thanks