Linear Regression gives NaN

Hi,

I am trying to do and plot a linear regression for the dataset attached. I tried following examples online, but I am getting NaN. This is my code and my data is attached.

clc
close all
clear all

%% Map
filename = 'TV_NYMA';
[num,string,vt] = xlsread(filename);
Year = num(:,1); 
County = string(:,2);
VOC = num(:,3);
NOx = num(:,4);
CO = num(:,5);
PM25 = num(:,6);
Lat = num(:,7);
Lon = num(:,8);
%%
standard_NOx = normalize(NOx);
standard_VOC = normalize(VOC);
figure
scatter(standard_NOx,standard_VOC)
title('NOx vs VOC')
xlabel('NOx emissions standardized')
ylabel('VOC emissions standardized')
%%
X = [ones(length(standard_NOx),1) standard_NOx]
b = X\standard_VOC
regression_line = [ones(size(standard_NOx,1),1) standard_NOx]*b

I was wondering what am I doing wrong.

Thanks.

[dpb]

I didn't download the data, but why not

b=polyfit(standard_NOx,standard_VOC,1);
yhat=polyval(b,[min(standard_NOX) max(standard_NOX)]);

or, if have one of Statistics or Curve Fitting Toolboxes, there are other higher-level routines as well...

Just out of curiosity, did you look at what was returned for the coefficients matrix? Was it NaN there, already? IF so, there's probaby a NaN in the mix in the data somewhere.

[John Ziggs]

There indeed was an empty row in the dataset which caused the NaN. Thank you for your help! I appreciate it.

Answer 1

[Star Strider]

There are 8 NaN values in those variables.

Eliminate them and it works —

T1 = readtable('https://www.mathworks.com/matlabcentral/answers/uploaded_files/625018/TV_NYMA.xlsx','VariableNamingRule','preserve')

T1 = 1803×8 table
    Year     County       VOC      Nox       CO      PM2.5     Lat       Long 
    ____    _________    _____    _____    ______    _____    ______    ______

    2010    {'BRONX'}    12.78    54.74    151.49    7.07     40.826    -73.92
    2013    {'BRONX'}     0.33     7.03      5.01    0.44     40.826    -73.92
    2013    {'BRONX'}      0.4      1.5      0.71    0.58     40.826    -73.92
    2014    {'BRONX'}     0.33    10.94      5.37    0.45     40.826    -73.92
    2015    {'BRONX'}     0.54    28.07      3.84    0.25     40.826    -73.92
    2014    {'BRONX'}        0        0         0       0     40.826    -73.92
    2014    {'BRONX'}     0.38      9.5      7.41    0.66     40.826    -73.92
    2011    {'BRONX'}     0.11     7.88      1.97       0     40.826    -73.92
    2013    {'BRONX'}     0.12    27.62      2.98    1.12     40.826    -73.92
    2012    {'BRONX'}      6.4    67.55     40.88    1.68     40.826    -73.92
    2014    {'BRONX'}     0.23    36.07      5.02    0.34     40.826    -73.92
    2015    {'BRONX'}     0.35     8.47      6.16    0.52     40.826    -73.92
    2016    {'BRONX'}     5.22    65.32     11.74       0     40.826    -73.92
    2011    {'BRONX'}     0.42     1.65      0.69     0.9     40.826    -73.92
    2013    {'BRONX'}     3.47    61.06     35.43    3.37     40.826    -73.92
    2013    {'BRONX'}     0.43     5.71      4.38    0.39     40.826    -73.92

% Nox_Nan = nnz(isnan(T1.Nox))
% VOC_NaN = nnz(isnan(T1.VOC))
T1 = T1(~[isnan(T1.Nox) & isnan(T1.VOC)],:)

T1 = 1795×8 table
    Year     County       VOC      Nox       CO      PM2.5     Lat       Long 
    ____    _________    _____    _____    ______    _____    ______    ______

    2010    {'BRONX'}    12.78    54.74    151.49    7.07     40.826    -73.92
    2013    {'BRONX'}     0.33     7.03      5.01    0.44     40.826    -73.92
    2013    {'BRONX'}      0.4      1.5      0.71    0.58     40.826    -73.92
    2014    {'BRONX'}     0.33    10.94      5.37    0.45     40.826    -73.92
    2015    {'BRONX'}     0.54    28.07      3.84    0.25     40.826    -73.92
    2014    {'BRONX'}        0        0         0       0     40.826    -73.92
    2014    {'BRONX'}     0.38      9.5      7.41    0.66     40.826    -73.92
    2011    {'BRONX'}     0.11     7.88      1.97       0     40.826    -73.92
    2013    {'BRONX'}     0.12    27.62      2.98    1.12     40.826    -73.92
    2012    {'BRONX'}      6.4    67.55     40.88    1.68     40.826    -73.92
    2014    {'BRONX'}     0.23    36.07      5.02    0.34     40.826    -73.92
    2015    {'BRONX'}     0.35     8.47      6.16    0.52     40.826    -73.92
    2016    {'BRONX'}     5.22    65.32     11.74       0     40.826    -73.92
    2011    {'BRONX'}     0.42     1.65      0.69     0.9     40.826    -73.92
    2013    {'BRONX'}     3.47    61.06     35.43    3.37     40.826    -73.92
    2013    {'BRONX'}     0.43     5.71      4.38    0.39     40.826    -73.92

standard_NOx = normalize(T1.Nox);
standard_VOC = normalize(T1.VOC);

X = [ones(length(standard_NOx),1) standard_NOx];
b = X\standard_VOC

b = 2×1

-0.0000 0.5010

regression_line = [ones(size(standard_NOx,1),1) standard_NOx]*b;

figure
plot(standard_NOx, standard_VOC, 'p')
hold on
plot(standard_NOx, regression_line, '-r')
hold off
grid

.

[John Ziggs]

Thank you for your help! I should've realized the empty row in the data would cause the NaN. Thanks again!

[Star Strider]

As always, my pleasure!