ksdensity function for pdf estimation

i feed some data to ksdensity but i got a gaussian pdf with peak greater than 1 how is that

Answer 1

[David Sanchez]

As you can see in the ksdensity documentation, depending n how you use the function, there is nothing abnormal with your result:

From documentation: Generate a mixture of two normal distributions, and plot the estimated inverse cumulative distribution function at a specified set of values:

x = [randn(30,1); 5+randn(30,1)];
yi = linspace(.01,.99,99);
g = ksdensity(x,yi,'function','icdf');
plot(yi,g);

Paste the code you are using in order to provide a good solution to your question.

[yasser]

let num =100

if true function [] =dist_vs_code_rate_non_negl_interf(num) %clear; disp('dist_vs_code_rate_non_negl_interf') n=6; %#of antenna fc=1700*1e6; %carrier freq for lte % BW=1.25*1e6; %operating freq range for narrowband lte could also be 2.5MHZ % fs=3*BW; %sampling freq c=3e8; %speed of light lambda=c/(fc); d=0.5*lambda; %inter antenna separation k=1; %#of usrs %num_pkts = 10; num_pkts=num noise_var_pre=1;

% obs_per=1; %observation period % t=0:1/fs:obs_per-1/fs; % no_samples=length(t); %is fs*obs_per

theta_test = [30]*pi/180; % Test vector of input azimuthiz theta values

A=(exp((transpose(-i*2*pi*d*[0:n-1])*sin(theta_test)/lambda))); %true steering mat for DOA

fft_size=128; %narrow band of lte

code_rate_pre=0.75; code_rate_post=0.5; code_rate_chk=0.5;

M=64;

no_samples_pre=.75*fft_size*log2(M)*code_rate_pre; no_samples_post=.75*fft_size*log2(M)*code_rate_post; no_samples_chk=.75*fft_size*log2(M)*code_rate_chk;

noise=randn(n,1.25*fft_size)*sqrt(1/2)+i*randn(n,1.25*fft_size)*sqrt(1/2); %Uncorrelated noise

loop=1;

for j=1:loop

%pre and negligable case

    for m=1:num_pkts

        sig=randi([0 1],no_samples_pre,1);
        data=transpose(sig);
        coded_data =convolution(data,code_rate_pre,10); %not built in fn in matlab code rate 3/4

        % Create a rectangular 64-QAM modulator System object with bits as inputs and Gray-coded signal constellation
            qamModulator = comm.RectangularQAMModulator(M,'BitInput',true);
        % Rotate the constellation by pi/4 radians
            qamModulator.PhaseOffset = pi/4;
        % Modulate and plot the data
            prechange = step(qamModulator, transpose(coded_data));

            pre_ofdm=fft_size*ifft(prechange,fft_size);

     %    Adding Cyclic Extension
           cext_pre=zeros(fft_size+0.25*fft_size,1);%ofdm symb is .25*fft_size(GI)+fft_size(data)
           cext_pre(1:.25*fft_size)=pre_ofdm(fft_size-0.25*fft_size+1:end);

           for k=1:fft_size
             cext_pre(k+.25*fft_size)=pre_ofdm(k);
           end

            X_pre=A*transpose(cext_pre)+noise;

            R_pre=X_pre*X_pre'/size(X_pre,2);

            b_pre(:,:,m)=R_pre;

    end
                                                                    %change code rate                                                                  
    for m=1:num_pkts
        sig_post=randi([0 1],no_samples_post,1);
        data=transpose(sig_post);
        coded_data =convolution(data,code_rate_post,10); %not built in fn in matlab code rate 1/2

        % Create a rectangular 64-QAM modulator System object with bits as inputs and Gray-coded signal constellation
            qamModulator = comm.RectangularQAMModulator(M,'BitInput',true);
        % Rotate the constellation by pi/4 radians
            qamModulator.PhaseOffset = pi/4;
        % Modulate and plot the data
            post_interf = step(qamModulator, transpose(coded_data));

            post_interf_ofdm=fft_size*ifft(post_interf,fft_size);

     %    Adding Cyclic Extension
           cext_post=zeros(fft_size+0.25*fft_size,1);%ofdm symb is .25*fft_size(GI)+fft_size(data)
           cext_post(1:.25*fft_size)=post_interf_ofdm(fft_size-0.25*fft_size+1:end);

           for k=1:fft_size
             cext_post(k+.25*fft_size)=post_interf_ofdm(k);
           end

            X_post_interf=A*transpose(cext_post)+noise;

            R_post_interf=X_post_interf*X_post_interf'/size(X_post_interf,2);

            b_post(:,:,m)=R_post_interf;

    end
                                                                    %chk phase

    for m=1:num_pkts

        sig_chk=randi([0 1],no_samples_chk,1);
        data=transpose(sig_chk);
        coded_data =convolution(data,code_rate_chk,10); %not built in fn in matlab code rate 1/2

        % Create a rectangular 64-QAM modulator System object with bits as inputs and Gray-coded signal constellation
            qamModulator = comm.RectangularQAMModulator(M,'BitInput',true);
        % Rotate the constellation by pi/4 radians
            qamModulator.PhaseOffset = pi/4;
        % Modulate and plot the data
            post_chk = step(qamModulator, transpose(coded_data));

            post_interf_ofdm_chk=fft_size*ifft(post_chk,fft_size);

     %    Adding Cyclic Extension
           cext_chk=zeros(fft_size+0.25*fft_size,1);%ofdm symb is .25*fft_size(GI)+fft_size(data)
           cext_chk(1:.25*fft_size)=post_interf_ofdm_chk(fft_size-0.25*fft_size+1:end);

           for k=1:fft_size
             cext_chk(k+.25*fft_size)=post_interf_ofdm_chk(k);
           end

            X_post_chk=A*transpose(cext_chk)+noise;

            R_post_interf_chk=X_post_chk*X_post_chk'/size(X_post_chk,2);

            b_post_chk(:,:,m)=R_post_interf_chk;
    end

%calc dist

    for m=1:num_pkts
            pre_post_dist(m)=(norm(10*log10(eig(b_post(:,:,m),b_pre(:,:,m)))));
            post_chk_dist(m)=(norm(10*log10(eig(b_post(:,:,m),b_post_chk(:,:,m)))));
    end

    [f1 xi]=ksdensity(pre_post_dist);

    [f2 xii]=ksdensity(post_chk_dist);

    plot(xi,f1);
    grid on;
    hold on
    plot(xii,f2,'r');
    legend('pre-post' , 'post-chk');
end
  end

Answer 2

[yasser]

please some one post an explanation i need it urgently