How do I implement Linear regression with leave-one-out cross validation in MATLAB?

I have a data set of 87 variables and 1 outcome where all are continuous. I need to use linear regression with leave-one-out cross validation to create a model/equation with prediction's accuracy, sensitivity, and specificity.

• Can I use Regression Learner app for this? If yes how to get quation from a created model in Regression Learner app?
• Can/Should I use cross validation to divide my data set and get the results?
• Is there any code sample on how to go about doing this? (I'm new to all of this)
• Should I be using stepwiselm or fitlm or glmfit/glmval? What's the difference and how do I choose?

Answer 1

[Amogh Bhole]

Hi,

In order to use Linear regression with cross validation you need to use fitrlinear, refer to this link for more details.

To apply leave one out cross validation use kfold keeping the value of k as the total number of observations in the data. Refer to this link.

You will find the answer to rest of the questions in the above links.

Answer 2

[Satadru Mukherjee]

Demo code of Implementation linear regression with leave-one-out cross validation in MATLAB

Note: I have tried to avoid the inbuilt functions to create the model or to cross validate or to calculate coefficient of determination , so that we can get the complete feeling out of the code:-)

clc
clear all
close all
warning off
data=readtable('Leave_One_Out.csv');
x=table2array(data(:,1))';
y=table2array(data(:,2))';
n=length(x);
predictions=[];
for p=1:n
    trainindex=setdiff(1:n,p);
    testindex=p;
    xtrain=x(trainindex);
    ytrain=y(trainindex);
    xtest=x(testindex);
    a=[];
    for i=1:length(xtrain)
        a=[a ; xtrain(i) 1];
    end
    c =a\ytrain';
    ytest = c(1)*xtest + c(2);
    predictions=[predictions ytest];
end
r=y-predictions;
r=r.^2;
rsq=(1-(sum(r)/sum((y-mean(y)).^2)))*100;
disp(rsq);