Is it possible to do a Missing completely at random (MCAR) test in multivariate data in matlab using, for example, the Little's MCAR test?

Is it possible to do a Missing completely at random (MCAR) test in multivariate data in matlab using, for example, the Little's MCAR test? I cannot find a function to do this type of test

Answer 1

[Animesh]

Hey @Guillermo Quintas

As far as I know, there is not any built-in function specifically for MCAR test (such as Little’s MCAR test) in MATLAB.

However, you can implement the test yourself by using various built-in functions and libraries in MATLAB for statistical and matrix operations.

Below is a basic approach to implement a simplified version of Little’s MCAR test in MATLAB (the actual computation may require more sophisticated handling when the pattern of missingness is complex).

function [chi2stat, df, p] = littles_mcar_test(X)      
% Listwise deletion to handle missing data 
X_complete = X(~any(isnan(X), 2), :);      
% Observed covariance matrix and mean vector 
S_obs = cov(X_complete, 'partialrows'); 
mu_obs = mean(X_complete, 'omitnan');   
% Expected covariance matrix under MCAR 
S_exp = cov(X, 'partialrows');   
% Degrees of freedom 
[n, m] = size(X); 
df = m * (m - 1) / 2;   
% Chi-square statistic 
chi2stat = n * (trace(S_exp \ S_obs) - log(det(S_exp \ S_obs)) - m);   
% P-value 
p = 1 - chi2cdf(chi2stat, df);  
end 

Hope this answers your query.

[Guillermo Quintas]

Thanks! that was very helpful!!