Dummy Variables in non linear regression models

Hello everybody!

i am quite a beginner in matlab and also in general in analyzing data and creating models.

Anyway I want to create a non-linear model that best fits my data.

Approximately:

ln Y= B0+B1X1+B2X2+ B3X3+Errorterm

It is suggested to work with the fitnlm function but I also have categorical variables and the documentation under Non linear Regression says :

"You cannot use categorical predictors for nonlinear regression. A categorical predictor is one that takes values from a fixed set of possibilities."

So can I create dummy variables so that I can use my categorical variables in my non linear regression model? Or what are other ways to create such a model?

Thanks for the answer and help

Dominik

Answer 1

[Apurvi Mansinghka]

Answer:

Hi Dominik,

I understand you want to handle categorical variable with nonlinear regression function fitnlm.

You can use dummyvar(group) function to get numeric representation for any categorical variable.

Example:

Suppose there is a dataset with a categorical variable 'Colours' that can take any of 3 values {'Red','Blue','Green'}

The dataset has 6 rows with following value for 'Colours':

Colours = {'Red';'Blue';'Green';'Red';'Green';'Blue'}; 

1.Create a dummy variable 'D' for the categorical variable

D = dummyvar(Colours) 

This gives the following result :

D = 6×3 
     0     0     1 
     1     0     0 
     0     1     0 
     0     0     1 
     0     1     0 
     1     0     0 

The columns in D correspond to the levels in Colours. For example, the first column of dummyvar corresponds to the first level, 'Blue', in Colours.

2. Now each column of D is a variable in your regression.

Create a non-linear model function with the additional variables

modelfun= @(k,x)(k(1)*x(:,1)+k(2)*x(:,2)+k(3)*x(:,3))... 
*(k(4)*x(:,4))*(k(5).x(:,5)); 

3. Set the parameter beta0 and create the model using fitnlm

beta0 = [-50 500 -1 500 -1]; 
mdl = fitnlm(tbl, modelfun, beta0)

Prefer to go through the tips section of the below link to understand best practices to handle categorical predicators with dummyvar:

https://www.mathworks.com/help/stats/dummyvar.html

[Dominik Gautschy]

Alright, it helped!I can work with this!Thanks for helping!

Greetings