How to generate scatter plot similar to attached scatter plot?

I want to generate scatter plot over laid by the equation of fit and R^2 value. Also the x and y axis lables as attached pdf plot. I am attaching the input data and plot and request you all to kindly suggest me how to do it in matlab. I would appreciate your kind suggestions.

Devendra

[dpb]

See <example> that shows a quadratic and its equation; simply follow those steps to fit your trendline after using scatter. Remember to set hold on before drawing the trendline to add to the same axis instead of clearing it.

Answer 1

[KSSV]

T = readtable('https://in.mathworks.com/matlabcentral/answers/uploaded_files/1458992/Test.csv') ;
plotregression(T.(1),T.(2))

If you don't have ML toolbox, you may follow:

T = readtable('https://in.mathworks.com/matlabcentral/answers/uploaded_files/1458992/Test.csv') ;
% plotregression(T.(1),T.(2))
x = T.(1) ;
y = T.(2) ;
R =  regression(x',y') ; 
p = polyfit(x,y,1) ;

figure(2)
scatter(x,y,[],'k','filled')
hold on
plot(y,polyval(p,y),'--b')

str{1} = sprintf('y=%0.2fx+%0.2f\n',p(1),p(2))  ;
str{2} = sprintf('R = %0.2f\n',R)  ;

text(2.8,10.5,str)

[Devendra]

Thank you very much. It worked. Only thing I want to reduce the gap between equation and Rsquare value. Once again thanks a lot for helping me.

Devendra

[KSSV]

Edited the code. Thanks is accepting/ voting the answer. :)

T = readtable('https://in.mathworks.com/matlabcentral/answers/uploaded_files/1458992/Test.csv') ;

x = T.(1) ;
y = T.(2) ;
R =  regression(x',y') ; 
p = polyfit(x,y,1) ;

scatter(x,y,[],'k','filled')
hold on
plot(y,polyval(p,y),'--b')

str = sprintf('y=%0.2fx+%0.2f\nR = %0.2f',p(1),p(2),R)  ;
text(2.8,10.5,str)

[Devendra]

Please also suggest me how to compute mean square error(MSE) and root mean square error(RMSE) in your scatter plot. I just want to get these values without overlaying over the plot.

Devendra

[dpb]

<regression> is documented as "Not recommended"; why such an animal was ever introduced is another Q? The better (and recommended) function would be fitlm which will produce all the desired results dirctly.

Nota Bene: The "R" value returned by regression is an unusual choice; it is NOT the normally-reported R^2 ("R-squared") which is the fraction of the total variance explained by the regression but the square root of R^2. Since the data in this case is mostly a blob rather than showing a linear relationship, the value of ~0.5 is pretty misleading; R-sq for these data is 0.22. That makes the lack of correlation much more apparent.

T = readtable('https://in.mathworks.com/matlabcentral/answers/uploaded_files/1458992/Test.csv');
T.Properties.VariableNames={'x','y'};   % make more convenient names
% IF going to use regression, at least use its results instead
% recalculating again later -- m,b are polyfit() coefficients
[R,m,b] =  regression(T.x.',T.y.')

R = 0.4712

m = 0.9168

b = 0.6520

scatter(T.x,T.y,[],'k','filled')
hold on
p=[m b];                            % slope, intercept for polyval()
plot(T.y,polyval(p,T.y),'--b')
str = sprintf('y=%0.2fx+%0.2f\nR^2 = %0.2f',p(1),p(2),R^2)  ;
text(2.8,10.5,str,'interpreter','Tex')

Now use recommended fitlim...

T.Properties.VariableNames={'X','Y'};   % make uppercase for pretty
f=fitlm(T,'Y ~ X')

f = 
Linear regression model:
    Y ~ 1 + X

Estimated Coefficients:
                   Estimate      SE       tStat       pValue  
                   ________    ______    _______    __________

    (Intercept)    0.65202     1.0723    0.60808        0.5442
    X               0.9168     0.1511     6.0677    1.3449e-08


Number of observations: 131, Error degrees of freedom: 129
Root Mean Squared Error: 2.43
R-squared: 0.222,  Adjusted R-Squared: 0.216
F-statistic vs. constant model: 36.8, p-value = 1.34e-08

It provides all the requested information directly

figure          % don't wipe out first
plotAdded(f)    % plot with regression line, model, bounds

You can adjust the plot as desired if don't want everything it puts on the plot by default and add the Rsq value if desired.