ansaurus

Question

Answer 1

+2 A:

You could try Kernel smoothing density estimate

Francisco 2010-04-08 05:52:40

+1 - You beat me to it! I just discovered this a few months ago, and use it regularly in place of histograms. Wikipedia has a fairly good explanation: http://en.wikipedia.org/wiki/Kernel_density_estimation

Doresoom 2010-04-08 14:01:49

Terrific ! ..... I never knew that such levels of 'smoothing' is possible !

Arkapravo 2010-04-09 03:29:43

Answer 2

+3 A:

If you pass two output parameters to HIST, you will get both the x-axis and y-axis values. Then you can plot the data as you like. For instance,

[counts, bins] = hist(mydata);
plot(bins, counts); %# get a line plot of the histogram

mtrw 2010-04-08 06:08:43

@mtrw : lovely one dude ! ...... you truly understood my problem !

Arkapravo 2010-04-08 07:47:51

@Arkapravo - glad it worked out.

mtrw 2010-04-08 21:03:39

@mtrw : Dude, it is great ..... but do you have any suggestions ....the curve look a bit jagged .... any smoothing would be the 'icing to the cake' !

Arkapravo 2010-04-09 03:40:19

@Arkapravo - You can try Francisco's suggestion (http://stackoverflow.com/questions/2597743/matlab-frequency-distribution/2597831#2597831). I've never done this myself, but it appears to try and estimate the probability density, rather than just count up a histogram. You should also make sure your bins are spaced such that each has the same number of possible data. However, I suspect that if you only have 500 observations, you're not going to get a statistically meaningful measurement whatever technique you use.

mtrw 2010-04-09 04:43:11

@mtrw : Francisco's method works great, however it 're-scales' the data elements, and I do not want that !

Arkapravo 2010-04-09 05:02:12

Sorry @Arkapravo, I don't know anything about the kernel smoothing density estimate, and I don't have the stats toolbox. Again though, with just 500 observations I'm not sure how good any pdf estimate will be.

mtrw 2010-04-09 07:06:34

@mtrw : Yeah you are right ! ..... but with more observations (say 10000) it works great !

Arkapravo 2010-04-09 07:19:50

@Arkapravo you can try to do a spline of the resulting curve (histogram) or you can try to scale your Kernel curve by this factor: Max(YourSample) / Max (KernelSmoothSample). It works because it appears that kernel smoothing tries to "hit" the max value of the sample at all cost. You can change the "width" parameter to obtain "more" or "less" smoothed curves. Hope it helps

Francisco 2010-04-11 18:55:13

ansaurus

tags:

views:

answers:

MATLAB : frequency distribution

related questions