Removing NAN values from the table and deleting it.

Hi

I have a table which is arrranged in susch a waym that it has one row of data and other row which contain NAN and so on, I want to get rid of NAN and aferwards deleting it.

Could you help me with this.

Answer 1

[madhan ravi]

see https://www.mathworks.com/help/matlab/ref/rmmissing.html introduced in 2016b

rmmissing(T) %deletes row containing nan where T your table

[madhan ravi]

According to your comment in the answer below this would give you the desired result:

T=table2array(T)
T(sum(isnan(T),2)==size(T,2),:)=[]
%OR
T(sum(isnan(T{:,:}),2)==size(T{:,:},2),:)=[] % in one step

[Shelender Kumar]

Thanks but It does not gives me what I wanted, as a example my data will look like below

NaN  NaN  NaN  NaN 
1.38E-01  -1.02E-05  -8.81E-02  -1.08E-05 
1.54E-01  NaN  -5.51E-02  NaN 
6.41E+01  7.83E-06  1.09E+02  3.83E-05 
NaN  NaN  NaN  NaN 

and answer should look like below

answer

1.38E-01  -1.02E-05  -8.81E-02  -1.08E-05 
1.54E-01  NaN  -5.51E-02  NaN 
6.41E+01  7.83E-06  1.09E+02  3.83E-05 

madhan ravi's reply: see the below comment it gives exactly the same result you want!

[madhan ravi]

>> 
format shorte
t=[NaN  NaN  NaN  NaN 
1.38E-01  -1.02E-05  -8.81E-02  -1.08E-05 
1.54E-01  NaN  -5.51E-02  NaN 
6.41E+01  7.83E-06  1.09E+02  3.83E-05 
NaN  NaN  NaN  NaN ];
T=table(t);
T=table2array(T)
T(sum(isnan(T),2)==size(T,2),:)=[]

T =

          NaN          NaN          NaN          NaN
   1.3800e-01  -1.0200e-05  -8.8100e-02  -1.0800e-05
   1.5400e-01          NaN  -5.5100e-02          NaN
   6.4100e+01   7.8300e-06   1.0900e+02   3.8300e-05
          NaN          NaN          NaN          NaN


T =

   1.3800e-01  -1.0200e-05  -8.8100e-02  -1.0800e-05
   1.5400e-01          NaN  -5.5100e-02          NaN
   6.4100e+01   7.8300e-06   1.0900e+02   3.8300e-05

>> 

[Shelender Kumar]

Thanks a lot but it does not work on my data file somehow

[madhan ravi]

Anytime :) ,it worked see the text file attached after running the code

[Shelender Kumar]

Thanks my mistake, I did not look carefully enough

[madhan ravi]

No problem :)

[Shelender Kumar]

could you help me understand this line, I do know about isnan but rest is not clear to me, thanks

T(sum(isnan(T),2)==size(T,2),:)=[]

[madhan ravi]

Ah I am bad at explaining things let me give it a try. isnan returns logical index which is summed across rows , so when the sum equals the column numbers of the matrix in which it's your table the entire row is deleted.

[Shelender Kumar]

Thanks, but I did not understand this part, could you explain me

size(T,2),:)

[madhan ravi]

size(T,2) gives you the number of columns in T

see size , matrix indexing and logical indexing to better understand what's going on

[Shelender Kumar]

Thanks a lot

[madhan ravi]

You're welcome , respond to your previous question in my comment

[Walter Roberson]

sum(isnan(T),2)==size(T,2)

is better expressed as

all(isnan(T),2)

[madhan ravi]

Thank you sir Walter , have to familiarise with any and all it‘s slightly confusing :)

[Talha Idrees]

not works

Answer 2

[ahmed nebli]

use this : (isnan(X)) = [] % X is the table

[Walter Roberson]

isnan() cannot be applied directly to a table. It can be applied to content of a table. In the case of a table that contains only numeric scalars you could do

X(any(isnan(X{:,:}),2),:) = [];

but you are probably better off using the new rmmissing() that madhan ravi mentions.

[Shelender Kumar]

Thanks , But I need something like this, let us say raw data look like this

 NaN  NaN  NaN  NaN 
1.38E-01  -1.02E-05  -8.81E-02  -1.08E-05 
1.54E-01  NaN  -5.51E-02  NaN 
6.41E+01  7.83E-06  1.09E+02  3.83E-05 
NaN  NaN  NaN  NaN 

answer

1.38E-01  -1.02E-05  -8.81E-02  -1.08E-05 
1.54E-01  NaN  -5.51E-02  NaN 
6.41E+01  7.83E-06  1.09E+02  3.83E-05 

[madhan ravi]

@Shelender see my comment in the above answer it will give you the desired result you want

[Talha Idrees]

not works

[Walter Roberson]

not works

Magik 8-Ball says:

Concentrate and ask again

Answer 3

[Francesco]

Following another question I found this code working really good:

Xnew=X((isfinite(X)));

The new array has no Nan inside.

[Walter Roberson]

This will not work for table objects.

Also, for 2D (or more) arrays, it has the side effect of unraveling the results into a column vector.

[Francesco]

It will work in this way, assuming that you have a X,Y,Z table you can delete the rows with NaN in X for example in this way:

T=[1,2,3;nan,2,3;1,2,3;1,2,3;nan,2,3]

T = 5×3

1 2 3 NaN 2 3 1 2 3 1 2 3 NaN 2 3

<mw-icon class=""></mw-icon>

<mw-icon class=""></mw-icon>

X=T(:,1) ; Y=T(:,2); Z=T(:,3);
Xnew=X((isfinite(X)));
Ynew=Y((isfinite(X)));
Znew=Z((isfinite(X)));

T_new = [Xnew,Ynew,Znew]

T_new = 3×3

1 2 3 1 2 3 1 2 3

<mw-icon class=""></mw-icon>

<mw-icon class=""></mw-icon>

You will have a table that doesn't have any NaN value in X

[Walter Roberson]

The question is about table objects.

T = array2table([1,2,3;nan,2,3;1,2,3;1,2,3;nan,2,3])

T = 5x3 table
    Var1    Var2    Var3
    ____    ____    ____

      1      2       3  
    NaN      2       3  
      1      2       3  
      1      2       3  
    NaN      2       3  

X=T(:,1) ; Y=T(:,2); Z=T(:,3);
Xnew=X((isfinite(X)));

Incorrect number or types of inputs or outputs for function isfinite.

Ynew=Y((isfinite(X)));
Znew=Z((isfinite(X)));
T_new = [Xnew,Ynew,Znew]

[Francesco]

Yes the principle is the same, if you have a table, for example in your case, you can call the columns using T.Column_name, in this case:

T = array2table([1,2,3;nan,2,3;1,2,3;1,2,3;nan,2,3])

T = 5x3 table
    Var1    Var2    Var3
    ____    ____    ____

      1      2       3  
    NaN      2       3  
      1      2       3  
      1      2       3  
    NaN      2       3  

X=T.Var1 ; Y=T.Var2; Z=T.Var3;
Xnew=X((isfinite(X)));
Ynew=Y((isfinite(X)));
Znew=Z((isfinite(X)));
T_new = array2table([Xnew,Ynew,Znew])

T_new = 3x3 table
    Var1    Var2    Var3
    ____    ____    ____

     1       2       3  
     1       2       3  
     1       2       3  

[Voss]

T = array2table([1,2,3;nan,2,3;1,2,3;1,2,3;nan,2,3])

T = 5x3 table
    Var1    Var2    Var3
    ____    ____    ____

      1      2       3  
    NaN      2       3  
      1      2       3  
      1      2       3  
    NaN      2       3  

T_new = T(isfinite(T.Var1),:)

T_new = 3x3 table
    Var1    Var2    Var3
    ____    ____    ____

     1       2       3  
     1       2       3  
     1       2       3  

Since the question only asked about NaNs, isnan might be a better function to use than isfinite, i.e.

T_new = T(~isnan(T.Var1),:)

T_new = 3x3 table
    Var1    Var2    Var3
    ____    ____    ____

     1       2       3  
     1       2       3  
     1       2       3