Smart PLS Flashcards
Structural Equation Modeling (SEM)
a family of statistical models that seek to explain the relationships among multiple variables
What can SEM be used for?
regression analysis, path analysis and factor analysis.
Research goals:
CB-SEM
PLS-SEM
Parameter-oriented
Prediction-oriented
Method:
CB-SEM
PLS-SEM
Covariance-based
Variance-based
Data assumption:
CB-SEM
PLS-SEM
Normal distribution
None
Good reasons to use PLS-SEM
- Estimation of complex models
- Integration of formatively measured constructs
- Working with small sample sizes
- Focus is on prediction
- Focus is on exploring new relationships
Not so good reasons to use PLS-SEM
- Focus is on exploring new relationships without having a hypothesized model.
- Working with small sample sizes (when the population is large)
Path model
a diagram that connects variables/constructs based on theory and logic to visually display the hypotheses that will be tested.
Correlation
linear relationship between two variables
range from -1 to +1
Covariance
- unstandardised form of correlation
- positive number leads to positive relationship
Reflective scale
changes in the latent variable directly cause changes in the assigned indicators
Formative scale
changes in one or more of the indicators cause changes in the latent variable
Stage 1:
Evaluation of the Measurement Model
Reliability models
- Indicator Reliability (Loading)
- Composite Reliability (CR) and Cronbach Alpha values
Indicator Reliability (Loading) and Composite Reliability (CR) and Cronbach Alpha values
Each indicator’s loading should be above 0.7
Convergent Validity
Average Variance Extracted (AVE) should be above 0.5, meaning that more than 50% of the variance is explained by the construct.
Discriminant Validity
Each construct should be distinct from others, often assessed using the Fornell-Larcker criterion or the Heterotrait-Monotrait ratio (HTMT).
Fornell-Larcker Criterion:
The square root of AVE of each construct should be higher than its highest correlation with any other construct.
HTMT:
Should be below 0.9 (or 0.85 for stricter threshold).
Reliability
the consistency or stability of a measurement instrument. A reliable instrument yields the same results under consistent conditions.
Internal Consistency:
Assesses the consistency of results across items within a test.
- Commonly used metrics:
Cronbach’s Alpha and
Composite Reliability (CR)
Indicator reliability
the degree to which an individual indicator (or observed variable) consistently and accurately measures the construct it is intended to measure. PLS-SEM= examining the loadings
loadings
measure of how strongly it correlates with its latent variable.
Validity
how well a test or measurement tool actually measures what it is intended to measure. It’s about accuracy.
- Convergent and Discriminant Validity
Convergent validity
how similar they are
Covergent Validity uses
Average Variance Extracted
AVE should be
greater than 0.5 = 50% of the variance in the indicators is explained by the construct
Discriminant validity
how different they are
Discriminant validity uses
Fornell-Larcker Criterion and HTMT
Fornell- Larcker
The square root of the AVE for each construct should be higher than its highest correlation with any other construct.
HTMT
The ratio should be below 0.9 (or 0.85 for stricter criteria).
Stage 2:
Check to see if there are any multicollinearity issues
Inner Collinearity
the correlation between the predictor constructs in the structural model.
- Assess for multicollinearity to ensure reliable path coefficient estimates.
Variance Inflation Factor (VIF)
measures the extent to which the variance of a regression coefficient is inflated due to multicollinearity with other predictors.
-It will show you which item has a problem
VIF should be
less than 5
Predictive relevance (R squared)
the proportion of variance in the dependent variables (endogenous constructs) that is explained by the independent variables (exogenous constructs).
- Evaluate how well the model explains and predicts the endogenous variables.
Model Fit
how well the model reproduces the observed data.
- Measure how well the model reproduces the observed data.
Standardized Root Mean Square Residual (SRMR)
SRMR is a measure of the average discrepancy between the observed and predicted correlations.
Significance and Path Coefficients:
Path coefficients represent the strength and direction of the relationships between constructs in the structural model.
- Check the statistical significance and practical relevance of the relationships between constructs.
Strong path coefficient
close to -1 or 1