iSixSigma

multiple regression in minitab

Six Sigma – iSixSigma Forums Old Forums General multiple regression in minitab

Viewing 10 posts - 1 through 10 (of 10 total)
  • Author
    Posts
  • #47887

    sree
    Member

    Hi
    Could someone please help me with this:
    I am trying to do a multiple regression in minitab. I have two input variables ( X1 and X2). So the regression equation in minitab is something like Y= 56+24X1+65X2. But I know that there is an interaction between X1 and X2. So I want minitab to include the interaction term X1X2  instead of just X1 and X2.
    Can anyone please let me know how to tell minitab to include X1X2 in the regression equation.
    Thanks and apprciate your help in advance!
     

    0
    #160194

    Erik L
    Participant

    Sree,
    Type Stat>Regression>Regression.  In the Predictors field use an asterisk between the two variables that you want Minitab to assess for interaction.  So, it’ll look like this:  X1 X2 X1*X2
    Regards,
    Erik

    0
    #160195

    sree
    Member

    Hi Erik,
    I tried it. Minitab says invalid characters when I click on the OK button.
    Thanks!

    0
    #160196

    Interaction in Minitab
    Participant

    If what Eric tells you doesn’t work go to calculate and create a new variable that multiplies x1 * x2, put it into a separate columns. Then run the regression by entering the new variable as an independent variable in the equation. The significance level of this new variable will tell you if the interaction is significant.

    0
    #160197

    Erik L
    Participant

    Sree,
    Got my platforms mixed up.  Unfortunately, you do have to manually create the interaction column and/or quadratic (etc.) if you want to incorporate them into the regression.  A little clunky.  Maybe they’ll address that with a future update….???
    Thanks,
    Erik

    0
    #160224

    BTDT
    Participant

    Sree:Look in the Minitab ‘help’ section and find the section on ‘specifying the model’. When the regression or General Linear Model dialog box comes up, you can click on ‘Model’ and specify the model as:X1|X2This should build the model as X1, X2 and X1*X2.The help section should show there is more than one way of specifying the interactions.Cheers, BTDT

    0
    #160594

    Savage
    Participant

    The two outputs to validate your stated claim that there is an interaction between X1 & X2 is the session window which shows a P-Value <.05 and the pareto chart included in the analyze DOE function. Based on the outputs, your final equation should include the two interaction factors provided there is an interaction.

    0
    #174862

    chinapanda
    Participant

    Maybe you can use stat > ODE > response surface > define……. to include your x1 and x2, then use stat > ODE > response surface > analyze … to inlude your y, and click “terms”, there is an item “linear”, change it to “interaction” or others, That’s it.

    0
    #174874

    Remi
    Participant

    Hai,
    What Chinapanda says does work (if you replace “ODE” with “DoE”).I use DoE often as an analysis tool when analysing regression or ANOVA models because it is user-friendlier with interactions. And I can include even quadratic terms if I use the RSM variant or 3-way interactions if I use the Factorial variant. You only have to be careful with outliers (choose -1/+1 settings different from min/max) but I always make a picture before I decide to analyse.
    The option where you put A|B or A*B in your model works for ANOVA but not for Regression.
     

    0
    #183991

    Murray
    Participant

    When you enter the data into minitab for multiple regression put X1, X2 and Y all into minitab. For the simple linear regression you would only put in Y and Either X1 or X2.
    In the minitab program, Go into Stat… Regression…Regression your resopnse is Y and your Predictors would be X1 and X2. from the regression equation you can see the response from each X1 and X2 on Y.

    0
Viewing 10 posts - 1 through 10 (of 10 total)

The forum ‘General’ is closed to new topics and replies.