PCA Graph of Data Comes Up Blank

I am running an example code from the Climate Data Toolbox and I am having issues plotting the graph of the principle component of my data. When I run the code, it is resulting in a blank graph. The data is the file attached. The following is my code.

clc 
close all
clear all
%% TS
filename = 'data';

S = dir(fullfile(filename,'*.txt'));
for k = 1:numel(S)
    fnm = fullfile(filename,S(k).name);
    mtx1 = load(fnm);
    mtx1 = normalize(mtx1);
    mtx = reshape(mtx1,72,144);
    TS{k} = mtx;
    TS{k}(TS{k} == -1000) = NaN;
end

TS_cat = cat(3,TS{:});
lat = (-90:2.5:90);
lon = (-180:2.5:180);

[Lon,Lat] = meshgrid(lon,lat);
figure
TS_cat = fillmissing(TS_cat,'linear');
imagescn(lon,lat,mean(TS_cat,3))
axis xy
cmocean thermal
xlabel 'longitude'
ylabel 'latitude'

%%
[eof_maps,pc,expv] = eof(TS_cat);

% Plot the first mode:
figure
imagesc(lon,lat,eof_maps(:,:,1))
axis xy image
cmocean('curl','pivot')
title 'The first EOF mode!'
t = length(pc(1,:));
%
figure
% subsubplot(3,1,1)
plot(t,pc(1,:))
box off
axis tight
ylabel 'pc1'
title 'The first principal components'

How would I fix this?

Thanks in advance.

Answer 1

[Chad Greene]

I think you're mixing and matching different ways of defining filenames. As a result, dir is unable to find the files you're looking for.

When I ran your code, this came up empty:

S = dir(fullfile(filename,'*.txt'));

The easiest solution is to navigate directly to the folder where all of your .txt files are, and do

S = dir('*.txt'); 

I was able to get around all of this by manually writing the filename in this line:

mtx1 = load('data.txt');

But that of course only loads the single example file you uploaded. As a result, pc only contains one entry, becauese there's only one timestep.

[Mike Zimmers]

In the portion for setting the filename, I have it set to the directory path.

filename = 'C:\Users\mike\Desktop\New folder\TS';
S = dir(fullfile(filename,'*.txt'));
for k = 1:numel(S)
    fnm = fullfile(filename,S(k).name);
    mtx1 = load(fnm);
    mtx1 = normalize(mtx1);
    mtx = reshape(mtx1,72,144);
    TS{k} = mtx;
    TS{k}(TS{k} == -1000) = NaN;
end

I checked what S gives and it gives this:

Is the portion "isdir" being 0 what you meant by S coming up as empty?

[Chad Greene]

Ah, okay, if the data's loading for ya, then that's not the issue.

After the loop, do you see any data in TS_cat? Check to make sure there's data in the first couple of slices by

imagesc(TS_cat(:,:,1))

and

imagesc(TS_cat(:,:,2))

[Mike Zimmers]

Yes, when I run

imagesc(TS_cat(:,:,1))

and

imagesc(TS_cat(:,:,2))

it gives me the following, respectively.

I noticed both maps are the same.

[Chad Greene]

They look similar, but they don't look exactly the same. Okay, so there's evidence that you do indeed have some data here. Are you sure there aren't any NaNs anywhere in the dataset? Check by

imagesc(sum(isnan(TS_cat),3))
colorbar

And make sure everything is zero.

[Mike Zimmers]

This is the image that I get when I run

imagesc(sum(isnan(TS_cat),3))
colorbar

I tried to see if changing

TS{k}(TS{k} == -1000) = NaN;

from NaN to 0 in the loop, makes any difference but it gives the same figure.

[Mike Zimmers]

I changed

TS{k}(TS{k} == -1000) = NaN;

to

TA{k}(TA{k} <= 0) = NaN;

in the loop and this is the figure that is produced: