Analysing mutiple structures of data

Hi,

I am fairly new to the MATLAB environment so please try to tolerate my incompetence :)

I have about 600 data sets of casts that measure water temperature and salinity through depth at different times in struct format.

I have written a script that calculates freezing temperature, depth, what depth freezing temperature and water temperature meet etc and it works fine on a single struct when loaded into the workspace.

But I want to analyse all 600 data sets and create a vector of depths where freezing point and temperature meet so I can plot this over time.

I have little to no programming experience but I'm thinking I can achieve this with a for or loop statement?

Any insight or help would be appreciated.

Cheers,

Ben

[Ben]

Hi,

Still a bit confused. I think the main thing that is confusing me is the "struct" format for the data. Each cast is in MATLAB format but when I open a cast each struct had the same. I took a screenshot to show what I mean: http://tinypic.com/r/2a4rbsy/5

It shows my script 'maxdepth' working on a single struct 'ctd' but I am not sure how to make it work on all the casts and then give me a vector of maxdepth and starttime without going through each one manually.

When I tried a for loop I couldn't get it to load anything.

Any ideas?

Cheers,

Ben

[Fangjun Jiang]

See update.

[Ben]

Hi,

Sorry I got sidetracked reading other ways to achieve what I am doing so I didn't see yours but I will also try it. This is how I have managed to do it.

% Loading data

clear all

files = dir('*.mat');

for i=3:length(files)

load(files(i).name)

% Setting variables

T = ctd.temperature;

S = ctd.salinity;

p = ctd.pressure;

Tp = ctd.pot_temp;

lat = str2double(ctd.latitude(6:11));

datetime = ctd.date_time;

% Calculating variables

X = (sind(lat/57.29578))^2;

gvar = 9.780318 * (1.0 + (5.2788e-3 + 2.36e-5 .* X) .* X) + 1.092e-6 .* p;

depth =((((-1.82e-15 .* p + 2.279e-10) .* p -2.2512e-5) .* p + 9.71659) .* p)./gvar;

Tf = fp(S,p);

% Max depth of supercooling

K = find(diff(sign(T-Tf))>0);

if K ~= 0

Tint = [T(K) T(K+1)];

Tfint = [Tf(K) Tf(K+1)];

Dint = [depth(K) depth(K+1)];

[x y] = curveintersect(Tint,Dint,Tfint,Dint);

Maxdepth = max(y);

Starttime = datenum(datetime);

elseif isempty(K) == 1

Maxdepth = 0;

Starttime = datenum(datetime);

end

end

The only trouble I am having now is that I am not sure how to get the vector of maxdepths. I know it is calculating them all because when I unsurpress the outputs it goes through and calculates them all.

Cheers,

Ben

[Fangjun Jiang]

That's good. You just need to pre-allocate the Maxdepth and then use Maxdepth(i) whereever you use Maxdepth.

N=length(files)-2;

Maxdepth=zeros(N,1);

for i=1:N

...

MaxDepth(i)=...

end

[Ben]

Hi, thanks for your answers btw, very helpful.

So I did what you suggested (I think):

N = length(files)-2;

maxdepth = zeros(N,1);

for i=1:N

K = find(diff(sign(T-Tf))>0);

if K ~= 0

Tint = [T(K) T(K+1)];

Tfint = [Tf(K) Tf(K+1)];

Dint = [depth(K) depth(K+1)];

[x y] = curveintersect(Tint,Dint,Tfint,Dint);

maxdepth(i) = max(y);

elseif isempty(K) == 1

maxdepth(i) = 0;

end

end

Now it seems to take a very long time to process the information (so long I have not been able to get an answer)

Do I have my code mixed up a bit? Can't seem to work out what I have done wrong.

Cheers,

Ben

[Fangjun Jiang]

It looks good for me, although I don't understand the rest of the code, for example, curveintersect(). You can use profile() to see where is the bottle neck.

[Ben]

I think I know what it is doing but not sure why,

It is writing each maxdepth for each one of the files onto the vector maxdepth over and over so by the end I end up with a vector which is (N,1) but full of the last value of maxdepth.

curveintersect is just my way of finding where T meets Tf

I'm working on it but I don't know why it isn't working

Thanks though,

Ben

[Ben]

Here is my entire code. Currently it is writing (N x 1) sized vectors of the maxdepth of 1st data file (i.e the same value N times) then replacing it with (N x 1) sized zero matrices then moving onto the next data file. Therefore at the end I am just left with a (N x 1) sized vector of the maxdepth last data file.

I think my for loops must be muddled up or something?

Can you see any obvious mistakes?

clear all

files = dir('*.mat');

for i=3:length(files)

load(files(i).name)

% Setting variables

T = ctd.temperature;

S = ctd.salinity;

p = ctd.pressure;

Tp = ctd.pot_temp;

lat = str2double(ctd.latitude(6:11));

datetime = ctd.date_time;

% Calculating variables

X = (sind(lat/57.29578))^2;

gvar = 9.780318 * (1.0 + (5.2788e-3 + 2.36e-5 .* X) .* X) + 1.092e-6 .* p;

depth =((((-1.82e-15 .* p + 2.279e-10) .* p -2.2512e-5) .* p + 9.71659) .* p)./gvar;

Tf = fp(S,p);

% Max depth of supercooling

N = length(files) - 2;

maxdepth = zeros(N,1);

for k = 1:N;

K = find(diff(sign(T-Tf))>0);

if K ~= 0

Tint = [T(K) T(K+1)];

Tfint = [Tf(K) Tf(K+1)];

Dint = [depth(K) depth(K+1)];

[x y] = curveintersect(Tint,Dint,Tfint,Dint);

maxdepth(k) = max(y);

elseif isempty(K) == 1

maxdepth(k) = 0;

end

end

end

Answer 1

[Fangjun Jiang]

A for-loop sounds good for your use case. If you provide a small example data, some might be able to help you figure out a way without the for-loop. But many examples have prove that using for-loop in new version of MATLAB is faster. Anyway, try it first. It's usually straightforward and it's easy to understand. If you have problems, show your code and people will help you.

If your function maxdepth(ctd) works fine, then what is your 600 data sets look like? Is it a struct array?

Try this to understand what is struct array. Use your ctd variable.

S_data=ctd; %make a copy
S_data(600).cast_number='600';
whos S_Data

It will tell S_Data is 1x600 struct, each one S_data(1), S_data(2),..., S_data(600) is a structure like ctd, you can do a for-loop

max_depth=zeros(600,1);
start_time=zeros(600,1);
for k=1:600
    [max_depth(k),start_time(k)]=maxdepth(S_data(k));
end

Now I noticed your .mat file, I think you need this:

max_depth=zeros(600,1);
start_time=zeros(600,1);
for k=1:600
    MatFile=sprintf('cast%3d.mat',k);
    ctd=load(MatFile);
    [max_depth(k),start_time(k)]=maxdepth(ctd);
end