What is a Covariance Matrix and how is it used in SEM (Structural Equation Modeling)

Поділитися
Вставка
  • Опубліковано 18 січ 2025

КОМЕНТАРІ • 15

  • @omaryahya1978
    @omaryahya1978 2 роки тому +1

    great explanation i have been searching for a simple videos like this that explain the concepts behind SEM method and eventually i found this ! so thank you very much sir!

  • @anupamk6725
    @anupamk6725 2 роки тому +1

    Clear and simple explanation.

  • @theresae194
    @theresae194 3 роки тому +2

    Hi Joel! This is a great explanation! Really, really helpful to understand the underlying mechanisms of SEM!

  • @musehime1142
    @musehime1142 5 місяців тому

    Thank you Professor!

  • @hfibona
    @hfibona 9 місяців тому

    Thank you for the video!

  • @Hashimhamza007
    @Hashimhamza007 3 роки тому +1

    Really insightful

  • @user-cat999
    @user-cat999 2 місяці тому

    Thats the explanation of observed cov matrix then how the explanation of estimated covarusnce matrix?

  • @lauracashman5473
    @lauracashman5473 2 роки тому

    Hi Joel, I was wondering if you have any resources regarding the minimum correlation strength required between variables to run an SEM?

    • @joelcollier9387
      @joelcollier9387  2 роки тому

      I don't think there is a "minimum" correlation necessary. Theory will determine your model more that retrospective correlations....Saying that, if you have really low correlations you are most likely not going to find significance in your model.

  • @rubymoran9359
    @rubymoran9359 2 роки тому

    Hello, Mr. Collier. May I ask if you have encountered a correlation between 2 constructs that is greater than 1 in SEM? I did. How do I explain this? Is it ok to consult?

    • @joelcollier9387
      @joelcollier9387  2 роки тому

      So you have what is called a "Heywood" case. This is an error in the analysis. You have multicollinearity between your two constructs. In essence, the constructs will not distinguish between one another because they are so similar. Your best bet is to drop one from the analysis and move forward.

    • @rubymoran9359
      @rubymoran9359 2 роки тому

      @@joelcollier9387 Thank you Sir for replying. May I ask one more question? I also got an R^2 that is greater than 1. Path coefficients can be greater than 1, right? But can R^2 be greater than 1?

    • @joelcollier9387
      @joelcollier9387  2 роки тому

      @@rubymoran9359 Standardized path estimates can not be greater than 1. Unstandardized path estimates can be greater than 1. With a R2 greater than 1, that says you are explaining more than 100% of the variance. Subsequently, This is an invalid result. My best guess is you have multicollinearity between constructs that is making your analysis go haywire.

    • @rubymoran9359
      @rubymoran9359 2 роки тому

      @@joelcollier9387 Thank you for your reply, Sir. I have two groups but this happens to only 1 group. 😔

  • @annacarolinapatricio6805
    @annacarolinapatricio6805 3 роки тому

    Brilhant