How to ignore columns which have only NaNs and move onto the next column for the mutli regression?

Hi,

I am trying to run a multi regression with 6 variables. Some of the variables data have columns which consist only out of NaNs. (all-NaN columns are not always the same for all variables)

How could I ignore columns which consist of all NaNs and move to the next column if this is a case for any of the variables?

Please see below my code for the regression:

numRegressors = 6;

numTickers = size(y,2);

numDays = size (y,1);

mat_betas=nan(numRegressors+1,numTickers); %beta coefficients

mat_tstats=nan(numRegressors+1,numTickers); %t-stats 

vec_rsqrds=nan(1,numTickers); %r-squares

mat_pval=nan(numRegressors+1,numTickers); %p-values

whichstats = {'beta','rsquare','tstat'};

for column = 1:size(y,2);

All_Factors = [a(:,column) b(:,column) c(:,column) d(:,column) e(:,column) f(:,column)];

all_stats = regstats(y(:,column),All_Factors,'linear',whichstats);

mat_betas(1:numRegressors+1,column)=all_stats.beta;

mat_tstats(:,column)=all_stats.tstat.t;

vec_rsqrds(:,column)=all_stats.rsquare;

mat_pval(:,column)=all_stats.tstat.pval;

end

Thank you in advance.

Answer 1

[KSSV]

You can check whether a number is nan or not using isnan. Read about it.

You can remove the NaN's from the column and then proceed.

% some random data 
data = rand(100,1) ;
data(randperm(100,30)) = NaN ;  % introduced NaN's

%%remove NaN's
data(isnan(data)) = [] ;

[Klio]

Thank you for your answer.

But by replacing NaNs the regression generates not precise results. Could you please possibly suggest how I could count the number of NaNs in the column and if it consists of only NaNs (or for example, has more than 200 NaNs) then move regression loop onto the next column?