How to add a line to the scatter?

Hello everybody,

I have a problem to find a code which plot the line on my scatter plot. I have upload a pictures of scatter and the example line which I have drawn manualy.

If someone can help, I would be glad.

Thank you

[Ameer Hamza]

Do you have x and y coordinate of the line? Or do you want to fit a line to the scatter points?

[Ilkin Abdullayev]

NO I just want to fit the line on the scatter plot which will show my track as I draw in the picture.

[Ameer Hamza]

Can you share the data points?

[Ilkin Abdullayev]

data points you mean x and y which i use to plot the scatter right?

[Ameer Hamza]

yes, You can attach that as a .mat file.

[Ilkin Abdullayev]

Here is the some part of the data.

x axis is distance, y is load. The result will not be the same as in the pictures that I upload because it was a data for different date.

Answer 1

[Ameer Hamza]

The data you shared is distributed vertically, so I estimated a vertical line using lsqcurvefit. You can use a similar method if you know an approximate equation that fits your model.

load('mathwork.mat');

x = @(p, y) p(1)*y+p(2);
p = lsqcurvefit(x, rand(1,2), data.Load, data.Distance);

scatter(data.Distance, data.Load)
hold on;
plot(x(p, data.Load), data.Load, 'LineWidth', 2);

[Ilkin Abdullayev]

Thanks a lot. Yes with this data it will be vertical because it is a very small part of whole. Normally, when I plot whole data as you may notice from the picture that I uploaded the graph is kind of fluctuation. Therefore do you think if I use:

x = @(p, y) p(1)*y+p(2);

p = lsqcurvefit(x, rand(1,2), data.Load, data.Distance);

I will get the fluctuated curve?

Well the code which I use is:

x = @(p, y) p(1)*y+p(2);

p = lsqcurvefit(x, rand(1,2), new3536.Load, new3536.Tunnelstation);

subplot(6,1,1);

scatter(new3536.Tunnelstation,new3536.Load);

hold on;

plot(x(p, new3536.Tunnelstation), new3536.Load, 'LineWidth', 2);

title('Load vs Chainage')

xlim([23435, 23436]);

I use subplot because I have 6 different plot that must be showed in one figure. Also I use for lop for these subplots. Does that can affect the workability of your code?

When I run these code it gives me :

lsqcurvefit stopped because the final change in the sum of squares relative to

its initial value is less than the default value of the function tolerance.

[Ameer Hamza]

The code I gave you will just make a straight line. If you want to make something else, then you will need to have an equation. If you don't have an equation, then you can use several other techniques. For example check following packages on FEX which do a similar thing

https://www.mathworks.com/matlabcentral/fileexchange/22470-lowess-locally-weighted-scatterplot-smoothing-for-linear-and-non-linear-data-enhanced

https://www.mathworks.com/matlabcentral/fileexchange/19195-kernel-smoothing-regression

If You have the curve-fitting toolbox, then you can also use smoothingspline option with fit(). Something like this

model = fit(new3536.Tunnelstation, new3536.Load, 'smoothingspline');
subplot(6,1,1);
scatter(new3536.Tunnelstation,new3536.Load);
hold on
plot(new3536.Tunnelstation, model(new3536.Tunnelstation), 'LineWidth', 2);

[Ilkin Abdullayev]

with curve fitting I gained the result. Thank you very much. I really appreciate.

[Ameer Hamza]

I am glad to be of help.