Robustness Checks using
Вставка
- Опубліковано 4 жов 2024
- In this video titled: Robustness Checks using #SmartPLS4, we will discuss how to perform structural model robustness checks that includes the assessment of Linearity, Endogeneity, and Heterogeneity.
The session is divided into three main sections, each addressing a crucial aspect of model assessment.
Linearity Assessment:
Learn how to check the assumption of linearity in Structural Equation Modeling (SEM) models. Understand the use of quadratic effects in SmartPLS, interpret P values, and confirm the linearity of relationships between variables.
Endogeneity Analysis:
Explore the assessment of endogeneity in your model using the Gaussian Copula approach. Follow a step-by-step guide to checking individual and combined relationships for endogeneity issues.
Heterogeneity Examination:
Gain insights into the significance of unobserved heterogeneity and its impact on model estimates. Discover the FIMIX procedure in SmartPLS for assessing heterogeneity. Understand the process of determining the optimal number of segments and interpreting model selection criteria.
By the end of this session, you'll be equipped with valuable knowledge and practical techniques to ensure the robustness of your structural model in SmartPLS4, covering linearity, endogeneity, and unobserved heterogeneity.
Contents
00:29 - Overview of the session
01:10 - Using quadratic effect in SmartPLS to assess linearity
02:30 - Confirming linearity when quadratic effects are insignificant
03:22 - Assessing endogeneity
09:07 - Introduction to unobserved heterogeneity and its importance
09:47 - Using finite mixture segmentation (FIMIX) in SmartPLS
10:30 - Determining the sample size for each segment
Do you have any questions, tips, or ideas? Let me know in the comments section below!
About:
Research With Fawad is a Research initiative to help young and early career research learn the research tools and techniques. It is an initiative to share research knowledge and help researchers learn to research the easy way.
Learn more about Research With Fawad at www.researchwi....
#SmartPLS4 #Robustness #Linearity #Endogeneity #Heterogeneity`
@Alfred from Seoul. Thank you Dr. Fawad for your detailed lectures. Your resources are life-saving.
Thank you.
This video is so helpful. Thank you!
Glad it was helpful!
Dear Dr
My calculated segment is 2.411 (sample size 164, gpower total sample size =68). decided to use 3 segments, the result shows:
AIC (segment 1), MDL-5 (segment 2)
AIC 3 (segment 1) & CAIC (segment 2)
AIC 4 (segment 2) & BIC(segment 2)
Can I consider there to be no issues of unobserved heterogeneity? Although the number of segments should be less than AIC and above MDL-5, which is not happening in this case?.
Tried to use 2 segments instead, but all criteria pointed to segment 1 only.
1 segment is fine.
I'm thinking if the sample less than 150, might no need for heterogenity test
Why?
Salam Dr. Fawad, For non-linearity test, if we choose p-value of 0.01 instead of 0.1 (one-tailed test) that we employed previously for the overall model bootstrapping. Is that acceptable? This is because two variables got 0.042 and 0.044.
WAS. Any thing less than 0.05 is referred to as significant.
@@researchwithfawad alright we just need to show the p-value of quadratic effect only, how about the direct linear effect above them? Also, is it necessary to do ramsey reset test afterwards? Or the quadratic effect in smartpls is sufficient?
Dear Prof, these days should be done after we validate the measurement model and before doing structural model? Thank you.
Sorry? I do not fully understand the question.
@@researchwithfawad Sorry, spelling mistake. It should be,
These robustness checks should be done after we validate the measurement model and before analyzing the structural model?
Or as a part of the structural model assement?
Do it before measurement model
@@researchwithfawad Thank you Professor.
I have curiosity that can we apply Gaussian copula on normally distributed to identify endogeneity? As one of the assumptions is "the endogenous regressor should exhibit sufficient nonnormality for the Gaussian copula approach to be applicable. Researchers need to confirm the nonnormality of the endogenous regressor through appropriate tests."
If you have the support for the proposition.
Do we need to check the linerity for HOC model if yes how
Yes you can check and the process remains the same.
well done!! but I have reported non linearity using SPSS especially one incorporates multi method such ANFIS with SEM
That is an option as well.
EXCELLENT. YOUR CONTENT IS VERY USEFUL. KEEP UP THE GOOD WORK. PLEASE LET ME KNOW IF I CAN HELP IN ANY MEANS@@researchwithfawad
Hi Dr. Fawad. May i know in what situation we need to conduct a robustness test?
It is recommended to be performed for each analysis.
Hi Dr. Fawad, could u demonstrate Ramsey Reset using Spss? Thank you
Salam Dr. Fawad, can I ask if there is non linear effect for some of the endogenous variables, how should we treat it or just remain as it is? For example FinTech adoption - online financing and income or FinTech adoption - online financing and subjective well-being. Because it can be explained that other factors mush have more weight on income besides online financing OR at a certain point of time, technology is not able to generate more income but for Zakat capital assistance, theres no cost involve except for internet connection or logistic cost to cybercafe to apply for free online financing.
WAS. Thanks for watching. How serious is the deviation. If other robustness tests are fine, you can have a cautious interpretation of results by mentioning the issue.
It seems like all failed but the robustness tests are just complementary test right?@@researchwithfawad
Yes