How declare for the matlab matrix that an value don't is considerable ??

How declare for the matlab matrix that an value don't is considerable?

I'm opening netcdf data and I want that the matlab ignore a specific value.

How do I?

Thanks! Carlos

[Carlos Batista]

I have a exemplo:

my matriz

a=[1,2,3,4,5]

I want ignore the value 2 for my matrix have only: (1,3,4,5) without the value "2"

How I do ?

Answer 1

[Titus Edelhofer]

You can remove elements from your matrix as follows:

a = [1 2 3 4 5];

% delete all two's:
a(a==2) = [];

Titus

[Carlos Batista]

thanks Titus!!!! It's ok, now!!!

Now I need to do the same thing for an larger matrix!!!

I will go test!

Thanks!

[Carlos Batista]

Titus

It's ok!!!

Well, now I want to ignore the 2 without erasing my matrix. How do I?

Answer 2

[dpb]

a=a(a~=2);

or

a(a==2)=[];

will replace a. If need to keep a for future reference, use the first form but assign RHS to another variable. NB: if the values aren't integers, floating point comparisons for exact values are not always reliable, you'll need to use "fuzzy" comparison or otherwise account for it.

[Carlos Batista]

thanks dpb!!! It's ok, now!!!

Now I need to do the same thing for an larger matrix!!!

I will go test!

Thanks!

[Carlos Batista]

dpb

Now I want to ignore the 2 without erasing my matrix. How do I?

thanks!

Answer 3

[dpb]

The size of the array has no bearing on the logic.

What I said, save it to another...

b=a(a~=2);

Or, there's the alternative of just using "logical addressing" directly without actually making a copy.

res_notwos=mean(a(a~=2));  % mean excluding elements equal 2

Or another alternative is

a(a==2)=nan;     % nan as placeholder
res=nanmean(a);  % Stat Toolbox has several such "NaN-aware" functions

The latter is particularly convenient when plotting as plot and friends simply ignore NaNs as well.

All depends on what you're objectives are as to which is the more convenient construct.

[Carlos Batista]

Titus

Very good, Titus! I had a matrix called data with 3D

date double (Are netcdf date)

Now I have a big doubt!

The Matlab ignores these NaN when he read? Matlab only considers the different NaN values? Can I do any math with these NaN in my series (matrix) that does not interfere with my calculations?

[Carlos Batista]

Titus

Very good, Titus! I had a matrix called data with 3D

date [180x89x1919 double] (Are netcdf date)

Now I have a big doubt!

The Matlab ignores these NaN when he read? Matlab only considers the different NaN values? Can I do any math with these NaN in my series (matrix) that does not interfere with my calculations?

[dpb]

No, Matlab normally does NOT ignore NaN with the exception of a few cases as previously noted. The Statistics Toolbox specifically has various functions that parallel the standard ones with the naming convention nanXXX that are coded to ignore NaN. Or, as also noted, plotting routines in general cleanly ignore them. Other than that, they're treated as any other value and will propagate thru the computations and likely sprinkle more around. So, it's a very specific solution for a very specific set of circumstances, NOT a general panacea. Again, only you can judge what's the most appropriate for you case--you've not given sufficient detail for us to know.

[Carlos Batista]

I have data in 3D

My matrix have this dimensions : 180x89x1919, where the 180 signify the number to longitudes, the 89 signify the number to latitudes and the 1919: serie time ...

my archive is an netcdf. This archive I have anomaly data and the date contours of continents (...)

When I open this my archive netcdf in matlab, my anomaly data comes together with the contour continents data. But, I need only to anomaly data (...)

I thought that using the "NaN" I could do the Matlab not consider the data contours of continents (are 32767) e consider only my data anomalies (...)

For a first test, ok (...) Very good!!!

The NaN replaced every 32767 (data contours continents). But, now these NaN are disturbing in my calculations (as advised) (...)

Now I need to make that these NaN don't participate of my accounts (...)

How I do?

[dpb]

Well, as before, you can use logical addressing to return the locations...

result=yourfunction(x(isfinite(x));

Now, the problem may be that this will return a linear index rather than the 3D since there isn't any regular array necessarily that the resulting points fall into and certainly not the original 3D array.

I can't envision precisely what the data would look like as described so not sure I've got any better answer.

[Carlos Batista]

How to make any math operation when there are NaN in data serial?

I'd like to make an test using in my data serial NaN...

For exemplo: I have 2 matrix (a and b) and I'd like of sum them!!!

Ex:

a=[1,2,3,NaN,5]

b=[1,2,3,8,5]

c=(a+b)

How can I to do?

[dpb]

Again, depends on what you mean by "sum" for the case...if you mean "ignore locations where value is NaN" then as noted before, if you have the Statistics Toolbox there are specialized funtions --

>> a=[1,2,3,NaN,5]; b=[1,2,3,8,5];
>> c=nansum([a;b])
c =
   2     4     6     8    10

If you don't have the Toolbox, you've got to write special code to do the same thing. Sometimes you can be "tricky" -- for addition,

>> a(isnan(a))=0;
>> c=a+b
c =
   2     4     6     8    10
>>

For multiplication, the magic value is '1'. In general you must basically process on an element by element basis taking action as desired when find the NaN. Often an anonymous function and arrayfun and/or accumarray can help immeasurably.

But, there is no one general answer.

Answer 4

[haseeb]

a(:,2)=[];

try this