Scatter Plot with different "markers" and "data labels"

a=[32.0 30.0 29.0 42.0 61.0 81.0]
b=[0.9 0.8 0.2 0.4 0.3 0.10]
c={'US'; 'GM'; 'SA'; 'IN'; 'EG'; 'BZ'}
scatter(a,b)
xlim([0 50])
ylim([0 1])
refline
grid 

I need a "scatter plot" with different "markers" and "data labels" (array in c). Also the refline should always start from (zero,zero)

[Adam]

doc refline

seems to have various options that take arguments.

[Charles Hendricks]

() Determine the cubic spline equations for the data set below:  () Plot f(x) as scatter plot, then add splines for each segment. a. (2pts) X and Y axes must be labeled properly x 0 100 200 400 600 800 1000 f(x) 0 0.653 1.0000 0.841 0.376 0.18813 0.08

Answer 1

[Adam Danz]

"I need a scatter plot with different markers"

The scatter() function only allows one marker definition so the data are plotted within a loop that iterates through a list of markers. The list of markers is replicated so that you never run out of markers in case the dataset grows, though that would result in duplicate markers if the number of points exceeds 13. Note that you could also change the color of the markers within the loop.

"...and data labels"

The data can be labeled either by a legend or by labeling the actual data points. The first block of code below shows how to use a legend to label the points. The secon block of code shows how to label the points on the plot.

"the refline should always start from (zero,zero)"

fitlm() is used to compute the slope of the regression line with an intercept of 0. refline() is used to draw the reference line.

See inline comments for details.

% Produce the data
a=[32.0 30.0 29.0 42.0 61.0 81.0];
b=[0.9 0.8 0.2 0.4 0.3 0.10];
c={'US'; 'GM'; 'SA'; 'IN'; 'EG'; 'BZ'};

% Define a sequence of symbols
syms = {'o' 's' '+' '*' 'x' 'd' '.' '^' 'v' '<' '>' 'p' 'h'};
% Replicate symbols if number of data points is larger than number of syms
syms = repmat(syms, 1, ceil(numel(a)/numel(syms)));

% Plot the data within a loop
hold on % important
h = gobjects(numel(a),1); 
for i = 1:numel(a)
    % Assign the label to the legend string using DisplayName
    h(i) = scatter(a(i),b(i),36,[0 0 1],syms{i},'DisplayName',c{i});
end

% Set axis limits
xlim([min(0, min(a)), max(a)])
ylim([min(0, min(a)), max(b)])

% Do the regression with an intercept of 0 and plot the line
linFit = fitlm(a,b,'Intercept',false);
refline(linFit.Coefficients.Estimate, 0);

% Add grid and legend
grid
legend(h)

Alternatively, you could lable the points directly on the plot using the labelpoints() function from the file exchange. Just add this line below to the end of the block of code from above (after downloading the labelpoints function).

% Label will be "North" of the datapoint with 0.1 spacing
labelpoints(a,b,c,'N',0.1)

[Muhammad Rahil Rafiq]

Magnificent !!!

[Adam Danz]

Happy to help!

[Muhammad Rahil Rafiq]

what if we plot the same data on "quarter polar plot"

[Adam Danz]

Is there something you're having trouble with adapting this to polar plots?

If so, share the code and describe what the problem is. I'd be glad to help.

Answer 2

[dpb]

Adam posted first but will add slightly different way to get to same place (sans some of the finishing niceties...)

Why scatter is so braindead re: markers by point is a wonder... :(

But, you can use plot instead; end up at similar thing in that multiple scatter calls also generate a handle for each call as do multiple lines in plot

N=nan(size(a));                     % dummy placeholder NaN array (see why later...)
hL=plot([N;a],[N;b]);               % create the plot w/ data, no markers, yet...
mkr={'o','+','*','x','s','d'};      % some markers...
set(hL,{'marker'},mkr')             % now put the markers on...
m=a.'\b.';                          % compute LS slope no intercept model via backslash operator
refline(m,0)
ylim([0 1])
grid

results in same figure as above w/o the labels yet.

The nan() array is needed to force plot to treat the two row vectors as separate columns by actually passing an array instead of just vectors--plot() will force a vector of either row or column to be treated as a single line instead of making a separate line by column as we need here to add a marker to each. NaN values are silently ignored so they don't show up as data; they're just placeholders to make plot do what is wanted.

[Adam Danz]

+1

Update: ignore this unless you want to be confused :D

I didn't know about this syntax m=a.'\b.';

It's mentioned briefly in the Tips section of mldivide but it differs so much from other Matlab syntax it makes me wonder if there are other similar syntaxes that treat a char array in this way. It's strange to me.

Also, m = b/a may be simpler and easier to understand.

[dpb]

"...if there are other similar syntaxes that treat a char array in this way."

??? a, b are doubles in OP's problem, they're (badly named it would appear) variables for his plot and the line. Not sure what the comment about char() arrays refers to at all.

mldivide is the general-purpose linear systems tool; as the Tips and Algorithms section indicate, it's got all sorts of tricks built into it for all kinds of things...before the geometric (but relatively recent) explosion of toolboxen and special-purpose functions or tools, it was the go-to. The relationship to mrdivide is there but I rarely remember it being of the age that many problems were only solvable by "rolling your own" design matrix and solving it instead of having something like fitlm that has the facility to not solve for the intercept. In the olden days, backslash and polyfit/polyval were about all there were.

The particular expression of a.'\b.' is simply another way of writing a(:)\b(:) to have column instead of row vectors as

>> a\b
ans =
         0         0         0         0         0         0
         0         0         0         0         0         0
         0         0         0         0         0         0
         0         0         0         0         0         0
         0         0         0         0         0         0
    0.0111    0.0099    0.0025    0.0049    0.0037    0.0012
>> 

not what we wanted at all...

But, back on the subject raised re: char() variables -- they're still just integers internally, only that the visual representation is encoded to match the ASCII equivalent. Internally, other than the limited range, "numbers is numbers"...

>> char(a.')\b.'
ans =
    0.0069
>>

Of course, one can't do

char(a.')\char(b.')

because b is floating point value that is non-integral and

>> double(char(b))
ans =
     0     0     0     0     0     0
>> 

which isn't the same numeric representation as for integer values fitting in range of uint8

[Adam Danz]

I see my mistake. At first glance I saw the line m=a.'\b.', and misinterpretted '\b.' to be a char array and some kind of unconvensional syntax of the backslash operator which is why I was surprised.

Now I clearly see that .' is merely a transpose without the compex conjugate operation imposed by ' alone. It's one of the few cases where the dot operator does not signify element-wise operations (or field-indexing of a structure). So, I'll blame it on faulty perceptual grouping ;)

Thanks, dpb!

[dpb]

Ah! Knowing what had written, I had failed to notice the possibility of misgrouping to have the perception of a char() string...

While not often for most, probably, the difference between " . '" and " .' " is profound if one has any complex variables around. Since much of my past consulting work dealt in the complex domain routinely, I soon learned to invariably use the dot version unless I specifically wanted/needed the other. Accidentally changing to the complex transpose vis a vis not if one is simply rearranging a variable's storage orientation is an insidious and often very difficult bug to find.

[Adam Danz]

I also use .' by default unless I'm transposing a cell array or some non-numeric datatype. I believe it was about a year ago I learned the importance of that. Still, when I see a.'\b.' my brain wants to see the char interpretation first. It's like the blue/gold dress.