Subtract one column from another unless NaN

I have a 100x4 double (data) with various NaN values throughout.

345 123 546 864
183 152 NaN NaN
842 NaN 238 945
NaN 283 965 NaN
392 105 932 NaN...
...

I am trying to subtract the first time point (that's not a NaN) from the last time point (that's not a NaN).

Here is the code I am using. It not only does not work, but makes me think there has to be an easier/more efficient manner of solving this.

for i = 1:100
    if isreal(data(i,4)) 
        if isreal(data(i,1))
            FirstMinusLast(i) = data(i,1) - data(i,4);
        elseif isreal(data(i,2))
            FirstMinusLast(i) = data(i,2) - data(i,4);
        elseif isreal(data(i,3))
            FirstMinusLast(i) = data(i,3) - data(i,4);
        else 
            FirstMinusLast(i) = 'N/a';
        end
    elseif isreal(data(i,3))
        if isreal(data(i,1))
            FirstMinusLast(i) = data(i,1) - data(i,3);
        elseif isreal(data(i,2))
            FirstMinusLast(i) = data(i,2) - data(i,3);
        else
            FirstMinusLast(i) = 'N/a';
        end
    elseif isreal(data(i,2))
        if isreal(data(i,1))
            FirstMinusLast(i) = data(i,1) - data(i,2);
        else
            FirstMinusLast(i)= 'N/a';
        end
    end 
end

[Totanly]

You mean A(:,1)-A(:,4) excluding NaN?

[Carver Nabb]

Yes, unless A(i,1) is NaN, then i would like to do A(i,2)-A(i,4). Assuming A(i,2) is real.

Answer 1

[madhan ravi]

help isnan

Answer 2

[Totanly]

c1=1 %or 2;
c2=3 %or 4
FirstMinusLast=data(:,c1)-data(:,c2); %if any NaN is there it will give NaN
data1=data(data>0);

Answer 3

[Totanly]

A=[345 123 546 864

183 152 NaN NaN

842 NaN 238 945

NaN 283 965 NaN

392 105 932 NaN]; %or make your own matrix

i=A>0;

A_new=A(i);

result=A_new(1)-A_new(length(A_new));