Controlling for parameters in Matlab regression models

Hi,

I wish to run a multiple linear regression analysis on some biomedical data, controlling for age and gender. How do I do this using the regress or glm functions?

_Barry

Answer 1

[TED MOSBY]

Hi,

To control for covariates like age and gender in a multiple‐linear regression, you simply include them as columns in your design matrix (for regress) or as terms in your model formula (for fitglm).

To use "regress" :

% Suppose you have:
%   Y        : n×1 vector of your primary response
%   age      : n×1 vector of age
%   gender   : n×1 vector coded 0/1
%   Xbio     : n×p matrix of your p predictors of interest

n = 100;
Y       = randn(n,1) * 10 + 50;            
Age     = randi([20, 70], n, 1);             
Gender  = randi([0, 1], n, 1);               
Bio1    = randn(n,1);                       
Bio2    = randn(n,1);                      

% Design matrix (intercept + age + gender + biomarkers)
X = [ones(n,1), Age, Gender, Bio1, Bio2];

% Multiple linear regression
[beta, betaCI, residuals, ~, stats] = regress(Y, X);

%   beta      : (p+3)×1 vector of coefficients
%   betaCI    : 95% confidence intervals for each coefficient
%   residuals : n×1 vector of residuals
%   stats     : [R^2, F-stat p-value, estimate of error variance, ...]

To use "fitglm":

T = table(Y, Age, Gender, Bio1, Bio2);

% Convert Gender to categorical
T.Gender = categorical(T.Gender, [0 1], {'Female','Male'});

% Linear regression
mdl = fitglm(T, 'Y ~ Age + Gender + Bio1 + Bio2');

disp(mdl)
disp('R-squared:'), disp(mdl.Rsquared.Ordinary)
disp('Coefficient table:'), disp(mdl.Coefficients)

Refer to these documentation links for more information on these functions:

https://www.mathworks.com/help/stats/regress.html?searchHighlight=regress&s_tid=srchtitle_support_results_1_regress#d126e1053417

https://www.mathworks.com/help/stats/fitglm.html?searchHighlight=fitglm&s_tid=srchtitle_support_results_1_fitglm

Hope this helps!