SEM Series (2016) 1. Introduction

James Gaskin

22 Apr 201605:46

Summary

TLDRThe speaker announces an update to their SEM series, reflecting changes in the field and the world. They will no longer cover tests for homoskedasticity and linearity due to current debates and journal practices. New additions include influentials tests, multivariate assumptions, and a multigroup analysis in Amos. The presenter corrects and clarifies common method bias tests, introduces a new mediation approach, and updates the validity tool. They also release an updated stats tools package, a new plugin, and a reference page for SEM. The series will cover a simplified model with three hypotheses, aiming to make the content more precise and literature-based.

Takeaways

📈 The speaker is launching a new series on Structural Equation Modeling (SEM) with updates reflecting changes in the field and the world.
🔍 Some outdated topics like tests for homoskedasticity and linearity are being removed due to current debates and less frequent requests from journals.
📚 New content will be introduced, including tests for influential observations and multivariate assumptions, as well as a test for multigroup analysis in Amos.
🛠 The speaker has clarified and corrected many aspects, improving the methodology for common method bias tests and multi-group comparisons using Chi-Square difference tests.
💾 An updated SEMands program is provided for easier identification of mediated or indirect effects, along with updates to the validity tool for better estimation of error and variance.
🔗 The speaker has created and updated plugins, including the CF plugin for the new Common Method Bias (CMB) approach and a new Amos A*B SEMands file.
📄 A new references page has been added to the wiki, compiling numerous references for SEM, with an invitation for contributions to expand the resource.
📊 The speaker emphasizes a shift towards more precise language and literature-based approaches in statistics, aiming to ground explanations and claims in research.
📊 The upcoming series will guide through an order of operations for SEM, starting with a simplified model for clarity and ease of understanding.
🎓 The data used in the series is from students at BYU, focusing on Excel usage, information acquisition, decision quality, and other variables, with demographic controls.

Q & A

What is the main reason for updating the SEM series?
-The main reason for updating the SEM series is that the speaker has learned a lot of new things, changed some approaches, and the world has also changed a bit, necessitating an update.
Which tests are no longer included in the updated SEM series?
-The tests for homoskedasticity and linearity are no longer included because there are ongoing debates about them, and most journals do not require them.
What new tests are being added to the updated SEM series?
-The new tests being added include tests for influentials, multivariate assumptions, a test for multigroup analysis in Amos, and the use of the multigroup models button in Amos.
How has the speaker improved the common method bias tests?
-The speaker has improved the common method bias tests by using literature-supported tests instead of the previously used ones that were not rigorously supported by literature.
What is the new mediation approach that the speaker mentions?
-The new mediation approach involves an s demand that the speaker has written, which can be downloaded and saved into the Amos area to help identify exact mediated or indirect effects.
What changes were made to the validity tool in the updated SEM series?
-The validity tool now uses exact error and variance for composite reliability estimation instead of estimated error, making it more accurate.
What updates were made to the plugins and tools mentioned in the script?
-The CF plugin has been updated to work with the new common method bias (CMB) approach, a new s demand file has been created, and the stats tools package has been updated. All these updates are available on the wiki.
Why did the speaker add a references page to the stat wiki?
-The speaker added a references page because it had been requested for years, and a participant online urged the speaker to do it, providing a bunch of references.
What is the purpose of the references page on the stat wiki?
-The purpose of the references page is to provide a collection of references for those doing SEM, which can be searched using the search function on the page.
How has the speaker's approach to statistics and language precision evolved?
-The speaker's approach has evolved to be more precise, with a focus on grounded literature-based approaches rather than pure logic, and the language used in explanations and claims has become more precise.
What is the model being tested in the updated SEM series?
-The model being tested is a simplified model based on data from students at BYU, focusing on how they use Excel, information acquisition, decision quality, usefulness, enjoyment, playfulness, atypical use, and controlling for demographic information like gender, age, and frequency of use.
How many hypotheses is the speaker proposing in the updated SEM series?
-The speaker is proposing three hypotheses in the updated SEM series for simplicity, although in practice, one might have multiple hypotheses.

Outlines

00:00

📊 Introduction to the Updated SEM Series

The speaker announces the launch of an updated series on Structural Equation Modeling (SEM), reflecting changes in the field and the world. They mention removing outdated tests, such as those for homoscedasticity and linearity, due to current debates and a shift in journal requirements. New content includes tests for influentials, multivariate assumptions, and a multigroup analysis in Amos. The speaker also discusses improvements in common method bias tests, now supported by literature, and introduces a new mediation approach with an s-demand file for Amos. They correct the validity tool to use exact error and variance, and mention updates to the CF plugin and the stats tools package. A new references page is added to the wiki, which is a compilation of numerous unorganized references accessible through search. The speaker emphasizes a shift towards more precise language and grounded, literature-based approaches in their explanations and procedures.

05:01

🔍 Overview of the Simplified Model and Hypotheses

The speaker outlines a simplified model for the SEM series, focusing on data from students at BYU regarding their use of Excel. The model includes variables such as information acquisition, decision quality, usefulness, enjoyment, and atypical use, along with demographic controls for age and frequency of Excel use. The speaker proposes three hypotheses for the model, emphasizing the simplicity for educational purposes. They note that in practice, researchers might have multiple hypotheses of various types, but for the series, they will test three distinct kinds. The speaker expresses hope that this approach will lead to a clearer understanding of SEM for the audience.

Mindmap

Keywords

💡SEM series

SEM stands for Structural Equation Modeling, a statistical technique used to analyze the relationships among observed and latent variables. In the context of the video, the SEM series refers to a collection of educational content focused on teaching and updating viewers on the latest methods and practices in SEM. The speaker mentions that the series is being updated to reflect new learnings and changes in the field.

💡Heteroskedasticity

Heteroskedasticity refers to the occurrence of non-constant variance in the residuals of a regression model. In the video, the speaker notes that tests for homoskedasticity are being removed from the series because of ongoing debates and the fact that most journals no longer require them, indicating a shift in statistical testing practices.

💡Linearity

Linearity in statistics pertains to the assumption that the relationship between variables in a model is linear. The speaker mentions that tests for linearity are also being dropped, suggesting a reevaluation of the importance of this assumption in current statistical analyses.

💡Influentials

Influentials, or influential observations, are data points that have a significant impact on the results of a statistical analysis. The speaker plans to include tests for influentials in the updated series, highlighting a focus on identifying and addressing data points that may skew results.

💡Multivariate assumptions

Multivariate assumptions refer to the conditions that must be met for multivariate statistical analyses to be valid. The speaker's intention to include tests for these assumptions in the series underscores the importance of ensuring data meets the criteria for robust multivariate analyses.

💡Multigroup analysis

Multigroup analysis is a statistical method used to compare different groups within a dataset to see if there are significant differences between them. The speaker mentions adding a test for multigroup analysis in AMOS, a software package for SEM, indicating an expansion of the series to cover more advanced analytical techniques.

💡Common method bias

Common method bias occurs when the measurement method itself introduces variance that can distort the results of a study. The speaker discusses updating the approach to testing for common method bias, moving from less rigorous to literature-supported tests, which shows a commitment to improving the validity of the analyses presented in the series.

💡Kai Square difference test

The Chi-Square difference test is used to compare the fit of two models in SEM to determine if one model fits the data significantly better than the other. The speaker's mention of this test in the context of multi-group comparisons suggests an emphasis on statistical techniques that can help researchers discern differences between groups.

💡Mediation

Mediation is a type of statistical analysis used to understand the indirect effect of an independent variable on a dependent variable through a mediator variable. The speaker introduces a new mediation approach and an sdemand file to help identify mediated effects, indicating a focus on clarifying and simplifying complex relationships in data.

💡Validity tool

The validity tool is a software utility used to assess the reliability and validity of measures in a study. The speaker notes improvements to the validity tool, such as using exact error and variance estimation, which reflects a move towards more precise and accurate statistical measures.

💡CMB approach

CMB stands for Common Method Bias. The CMB approach is a method for assessing and controlling for common method bias in research. The speaker's update to the CF plugin to work with the new CMB approach shows an evolution in the tools and methods used to ensure research validity.

Highlights

Introduction of a new SEM series with updates based on learnings and changes in the world.

Removal of tests for homoskedasticity and linearity due to current debates and less frequent requests from journals.

Addition of tests for influentials and multivariate assumptions.

Introduction of a test for multigroup analysis in AMOS using the multigroup models button.

Clarification and correction of many things, including better ways to do common method bias tests.

Use of literature-supported tests for common method bias instead of loosely supported ones.

Introduction of a new mediation approach with an SEM demand file available for download.

Updates on the use of Baron and Kenny's approach in mediation analysis.

Fixing the validity tool to use exact error and variance for composite reliability.

Update of the CF plugin to work with the new common method bias approach.

Creation of a new SEM demand file and update of the stats tools package.

Addition of a references page to the wiki, fulfilling a long-requested feature.

Invitation for users to contribute references for the SEM field.

Emphasis on more precise language and grounded, literature-based approaches in statistics.

Plan to walk through the order of operations as outlined on the general guidelines page.

Testing of a simplified model using the same variables and data as the previous series for consistency.

Proposing only three hypotheses for the sake of simplicity in the new series.

Description of the model focusing on student use of Excel, information acquisition, decision quality, and other variables.