plot too many lines

Why does MATLAB plot so many dashed lines when I am plotting just one for the 95% confidence intervals?

xd = bindex_total2_nonan(:,9);    % depth
yd = bindex_total2_nonan(:,6);    % pos

% poly fit
[pd,Sd] = polyfit(xd,yd,1);
[y_fitd,deltad] = polyval(pd,xd,Sd);
conf1 = y_fitd+2.*deltad;
conf2 = y_fitd-2.*deltad;

figure('Renderer', 'painters', 'Position', [10 10 900 600]);
scatter(bindex_total2(:,9),bindex_total2(:,6),'+','k');
hold on;
plot(xd,y_fitd,'color','k')
plot(xd(1:2:54),conf1(1:2:54),'r--')
plot(xd(1:2:54),conf2(1:2:54),'r--')

[Geoff Hayes]

Elias - what are the dimensions of the inputs to your calls to plot? Are the xd and y_fitd column or row arrays, or are they matrices? If matrices, that may explain why you are seeing many dashed lines...

[Walter Roberson]

We see from the first line that xd must be a column vector, and yd as well. y_fitd is the same length as xd generated from the fitted line.

[Walter Roberson]

I notice you never turn "hold off". If you were to run this code several times, all of the runs would get plotted together.

[Elias de Korte]

deltad, xd and y_fitd are both 54x 1 vectors, but I think it has something do do with these double values in the deltad output

deltad =

   25.9672
   25.9672
   25.9672
   25.4818
   25.4818
   25.7058
   25.7058
   25.7416
   25.7416
   25.3641
   25.3641
   25.3641
   etc.
   

[YT]

I was also looking at this problem, and if it's not the dimensions of xd/yfit that causes the problem maybe its indeed like @Walter Roberson suggested. Also maybe its a good practice to close figures (close all;) and clear workspace variables (clear;) at the start of the script.

[Elias de Korte]

I have used clear and close all for every run and this does not solve the problem. I have added hold off as wel, without result

Answer 1

[Walter Roberson]

Try

xu = unique(xd);  %unique and sorted
[y_fitd,deltad] = polyval(pd,xu,Sd);

and then

plot(xu, y_fitd, 'color','k');

I suspect that you will not need to skip every second point for the other plots.

[Elias de Korte]

yes, unique values solves the problem visually. But the problem is this:

for x- values i have 54 unique values, for example values concerning 3 months within the year. For the other y-variable i have only got yearly values, which is why some of the same y-values are assigned to different x-values. The solution would be to average the x-values for each year I guess.

[Elias de Korte]

Now i expected a linear confidence interval, but this is not the case. I can't find how to change the settings correctly

Answer 2

[Rik]

You should not use clear on its own. If you want to, you can use

clear variables

As to your actual problem, is your xd not sorted? That would cause the line to jump back and forth, making it look like you have a lot of different lines.

[YT]

Can you clarify why "you should not use clear on its own"? Isnt typing every specific variable you want to clear from your workspace much more troublesome than just using clear;?

[Rik]

Maybe I worded it a bit too strongly, but when people read the advice to use clear, they will often understand it to use this

clc,close all,clear all

This means that not just the variables are cleared, but also loaded functions and clases, slowing code down. You can read the options in the doc for clear.

Literally typing

clear variables

seems to do the same as only using clear, but this way you show your intent as well preventing misunderstandings.

Also, you can use wildcards and/or regular expressions to specify groups of variables you might want to clear.

[YT]

Ah okay good to know. Thanks for the clarification.