Want to fit linear curve on my data.

I have x, y data

 x=[0;0.100000000000000;0.200000000000000;0.300000000000000;0.400000000000000;0.500000000000000;0.600000000000000];
 y=[4.67178152947921e-06;4.67353333624452e-06;4.70560728038426e-06;4.74873086195845e-06;4.77333265701103e-06;4.84630647442201e-06;4.87015810633671e-06];

I want to plot x vs y and want to y-axis in log scale

 plot(x,y)
 set(gca,'YScale','log')
 hold on

Note: x data starts from 0

Now I want to fit a line and show the slope of that curve fitting line + original curve

          p=polyfit(x,(y),1);
          q=polyval(p,x);
          plot(x,q).

It seems to be not right because the fit line isn't straight ( it likes power fit or exponential) . Note log scale ( not log(data))

Please helps. Thanks

[the cyclist]

It would be tremendously helpful if you included a small sample of your data that exhibits the problem, so that we could run your code and see the results. Your statement that it "seems to be not right" is not quite enough.

[Daniel Shub]

Your x and y are not the same size ...

Answer 1

[Daniel Shub]

I don't like working on log scales, I would rather take the log transform of my data

x=[0;0.100000000000000;0.200000000000000;0.300000000000000;0.400000000000000;0.500000000000000;0.600000000000000];
y=[4.67178152947921e-06;4.67353333624452e-06;4.70560728038426e-06;4.74873086195845e-06;4.77333265701103e-06;4.84630647442201e-06;4.87015810633671e-06];

xx = x;
yy = log(y);

plot(xx, yy, '*');
lsline;
p = polyfit(xx, yy,1);
text(max(xlim), max(ylim), ['Slope: ', num2str(p(1))], 'HorizontalAlignment', 'Right', 'VerticalAlignment', 'Top')
set(gca, 'YTickLabel', 10^6*exp(get(gca, 'YTick')))

[Phong Pham]

But I Really need the log scale on y. I applied the your code and it does not look good since the plot x, y does not look linear curve so the fit line is way off.

[Daniel Shub]

Opps, I missed that. I edited the code.

[Phong Pham]

I dont know why but it has something weird to me. First, the y-scale has number likes 0.00344 not 10^-3 , etc .. ( log scale). Second, the slope is negative, the fitting line is straight up so it should give positive slope.

Thanks

[Daniel Shub]

I got the output of polyfit wrong (I thought the first coefficient was the intercept). I changed the code. I also changed the scaling of the ticks. You can set them to be whatever you want.

Answer 2

[Jan]

When you want a line in the logspace diagram, you need an exponential fit on the data. Or build the log of the data at first and fit the line afterwards.