Multiple regression with categorical variables

Hi, I'm new to Matlab sorry if my question is silly. I have dataset consists of 100 rows and 10 column which are Age, Diastolic, Gender, Height, systolic, LastName, Weight, Smoker, Location, SelfAssessedHealthStatus. I need to create a linear regression to predict systolic based on Age, Gender, Height, Weight, Smoker, Location, SelfAssessedHealthStatus. the problem for me is I have 3 categorical variables I'm not sure about how to deal with them in right way. belew is my try. can you please suggest to me how to deal with them..

if true
  load ('patients');
patients= table(Age, Gender, Height, Location, SelfAssessedHealthStatus, Smoker, Weight);
patients.Gender = nominal(patients.Gender);
dv = dummyvar(patients.Gender);
patients.Location = nominal(patients.Location);
dv1 = dummyvar(patients.Location);
patients.SelfAssessedHealthStatus = nominal(patients.SelfAssessedHealthStatus);
dv2= dummyvar(patients.Location);
x=[Age dv Height  Weight Smoker dv1 dv2];
y= Systolic;
ml=fitlm(x,y)

end

Answer 1

[Sid Jhaveri]

In "fitlm" function you can specify which variables are categorical. For more information on how to achieve this, I would suggest to refer the documentation example given in the link below:

https://www.mathworks.com/help/stats/fitlm.html#bt0clcj-1

[Tanvir Kaisar]

Hello, I have followed the suggested link. I have 2 questions:

1. Will it work when there are more than 2 levels of category? (Say for sex - male, female and transgender)

2.Will this approach also work for multinomial logistic regression mnrfit function? If it won't then how should I approach to deal with the categorical variables in case of mnrfit?

Answer 2

[Ebby Thomas]

what i believe is that the following code should work for you as per the documentation.

linearmodel = fitlm(patients,'ResponseVar','Weight','PredictorVars',{'Age', 'Gender', 'Height', 'Location', 'SelfAssessedHealthStatus', 'Smoker'},'CategoricalVar',{'Gender','Location','SelfAssessedHealthStatus'})

However, I am interested to know about your interpretation s of the result. Please share your findings as well..

Ebby