Plotting a curve with just x values

I want to fit a curve to a vector of x values that I have. I have a vector of x values which contains the frequency of occurence of each value. I mean if the vector is like this:

[3,3,4,5], it means that the height over x = 3 must be double than over x = 4 and x = 5. Now I want to plot the best curve fitting this vector.

My idea was to sort the vector increasingly and count the number of occurrences for each x value and plot the curve but even then I don't know how to make it a continous function.

Thanks!

[darova]

Can you create a simple drawing to explain more?

[Amir mo]

suppose we have this vector [0,0,1,2,3,2,-1,2] what i want to plot is a curve like this. The point is that I have the values for a limited range (many points but still limited) but i want to predict the function over the whole x axis

[dpb]

"i want to predict the function...over the whole x axis"

What does that mean, specifically? What do you expect a functional form to be? Extrapolation is an extremely risky proposition.

Answer 1

[dpb]

Something like

x=[3,3,4,5];
n=histc(x,unique(x,'stable'));
v=[1:numel(n)];
m=fitlm(v,n);
bar(n);
xlim(xlim+[1 -1]*0.5)
hold on
line(v,feval(mdl,v),'color','r')
set(get(gca,'YAxis'),'TickLabelFormat','%.1f')

maybe?

v=[0,0,1,2,3,2,-1,2];
x=unique(v);
n=histc(v,x);
stem(x,n)
xlim([x(1) x(end)]+[-0.25 0.25])
ylim([0 max(n)+0.1])
xticks(x)
yticks([0:max(n)])

results in

This idea of predicting this outside the range would appear to be a task for the Crystal Ball Toolbox that has yet to be released.

ADDENDUM: Mayhaps I was reading the original Q? too literally on the expectations...what about

xq=linspace(x(1),x(end));
ns=interp1(x,n,xq,'spline');
nc=interp1(x,n,xq,'pchip');
hold on
hL=plot(xq,ns,'-r',xq,nc,'-c');
ylim([0 ceil(max(nn))+0.1])
yticks([0:ceil(max(nn))])
legend('Data','Spline Interpolant','pchip Interpolant')

which results in

?

[Amir mo]

Thanks a lot!