In my PLS-SEM model most values r insignificant with negative paths. I assumed that because of extreme outliers having that problem. Can I delete only those questions? How am I really deal with that outliers?
Dear Halima,
I suggest that you follow these references to address robustness and leveraging outliers in PLS-SEM:
- Debruyne et al., " Robustness and Outlier Detection in Chemometrics ", 2006 -
Article Robustness and Outlier Detection in Chemometrics
- Hubert et al., " High-Breakdown Robust Multivariate Methods ", 2008 - https://arxiv.org/pdf/0808.0657.pdf
- Zhong et al., " Bias and Efficiency in Structural Equation Modeling: Maximum Likelihood Versus Robust Methods ", 2011 -
Article Bias and Efficiency in Structural Equation Modeling: Maximum...
- Yuan et al., " Robustness of Fit Indices to Outliers and Leverage Observations in Structural Equation Modeling ", 2013 - https://s3.amazonaws.com/academia.edu.documents/43664850/Robustness_of_Fit_Indices_to_Outliers_an20160312-26853-1fnm5cm.pdf?AWSAccessKeyId=AKIAIWOWYYGZ2Y53UL3A&Expires=1531906508&Signature=XjKZnobigXvFKPzplKwZFWlr0uk%3D&response-content-disposition=inline%3B%20filename%3DRobustness_of_fit_indices_to_outliers_an.pdf
- P.J. Rousseeuw, "Chapter 16 - Robust Estimation and Identifying Outliers", 1990 - https://wis.kuleuven.be/stat/robust/papers/publications-1990/rousseeuw-robustestimation-handbookengineersscient.pdf
Regards
You can do Winsorization to minimize the effect of outliers. You don't have to delete the sample and the sample size will remain the same.
first variable extreme or outliers is determined by a distance from the observed values which is not usually close to the center of gravity. We have several graphs that illustrate, it is enough to remove the individuals possessing this character far removed from the real value. Once again it takes mastery of the field to effectively identify if it really is an outlier. There is the suppression technique. In the case of several variables, it would be better to restart a survey or a statistical survey to better analyze the data ...
Dear Dr Asalm
https://www.youtube.com/watch?v=WJuB0vZp6w4...
I followed above link to winsorization but still have outliers (extreme).
Would u please help me to send a helpful link as how to winsorize in SPSS?
Kind Regards
Halima
How many percentile levels have you used in winsorization? you can do winsorization at 5 and 10% percentile levels and it should minimize the presence of extreme outliers.
Dear Halima,
Here are 2 references of interest to help you in detecting and replacing multivariate outliers in a score space context generated by PCA , based on winsorization (by Huber & Hampel):
- Tvarlapati et al., " A METHOD OF ROBUST MULTIVARIATE OUTLIER DETECTION AND REPLACEMENT ", 2000 - https://ac.els-cdn.com/S1474667017386135/1-s2.0-S1474667017386135-main.pdf?_tid=3dfb5f70-f6dc-4b96-a54e-48cf472e8af8&acdnat=1532005411_2b6ca819634cbad6c56c57cb577ca716
- Dixon et al., "Approximate Behavior of the Distribution of Winsorized t (Trimming/Winsorization 2)", 1968 - http://booksc.org/book/26751902/49e697
Regards
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Sampling Design
- Sample Size