Linear regression : How to take into account the quadratic term only and not the linear term ?

Hello everyone,

I am currently working on a linear regression using the fitlm tool. I collect my datas from a table name "newT2" that has 3 columns.

Here is my code :

modelaction = 'Var1 ~ Var2 + Var3^2';
mdlaction = fitlm(newT2, modelaction);

The problem is that the regression is done according this equation : 'Var1 ~ Var2 + Var3+ Var3^2'. But it's not what I asked, I don't want the linear term Var3.

In fact, if I run with the equation 'Var1 ~ Var2 + Var3^3', the regression is done according : 'Var1 ~ Var2 + Var3+ Var3^2 + Var3^3' and so on.

Thanks for your help !

[Torsten]

If nothing helps, square Var3 in newT2 and use a linear model:

modelaction = 'Var1 ~ Var2 + Var3_squared';

Answer 1

[Star Strider]

Suppress the intercept term and the linear ‘Var3’ by negating them —

newT2 = array2table(rand(7,3))                              % Create Data Table

newT2 = 7×3 table
     Var1       Var2        Var3  
    _______    _______    ________

     0.9339    0.56678      0.7716
    0.86767    0.18461    0.039493
    0.95488    0.51097     0.95676
    0.53864     0.8246     0.86486
    0.23089    0.41095     0.76704
    0.93338    0.12252     0.19638
    0.91871    0.38145      0.6025

modelaction = 'Var1 ~ - 1 + Var2 - Var3 + Var3^2';          % Specify Model
mdlaction = fitlm(newT2, modelaction)

mdlaction = 
Linear regression model:
    Var1 ~ Var2 + Var3^2

Estimated Coefficients:
              Estimate      SE       tStat      pValue 
              ________    ______    ________    _______

    Var2        1.7668    1.4213      1.2431    0.26895
    Var3^2    -0.38241    1.2132    -0.31521    0.76534


Number of observations: 7, Error degrees of freedom: 5
Root Mean Squared Error: 0.572

.