Hi All,
Please help me out. I am a honours student doing my thesis with CFA. It's a single factor scale.
Preliminary:
*No missing data
*Checked normality of each item, some z-score outside |1.96| but did not exceed |3.29|, so retain in the dataset.
*Two multivariate outlier, p1 in CFA output =
CFA (Confirmatory Factor Analysis) is used to test whether a hypothesized factor structure fits the observed data.
Multiple Group CFA extends CFA to compare factor structures across different groups (e.g., genders, age groups) to assess whether the same factor structure is valid for each group.
Both methods help in validating measurement models and ensuring consistency across different populations.
Hello Sir hope you well..
as i know these analysis well so i will tell you somthing according to my point of view..
Q1: Did I miss out anything in my analysis?
Yes, it seems that you may have missed some potential issues with non-normality. Although you checked for normality and removed two multivariate outliers, the multivariate kurtosis (12.80) and condition index (4.21) suggest that the data may still be non-normal. This could potentially impact the results of your confirmatory factor analysis (CFA). You may want to consider using robust methods or transformations to address non-normality in future analyses.
Q2: Can I conclude the model is correct with 1901 bootstrap samples?
Not necessarily. Although the Bollen-Stine bootstrap output indicates that the model fits better in 1901 bootstrap samples, the p-value of 0.050 suggests that the null hypothesis that the model is correct cannot be rejected at the 0.05 level. This means that the model may not be significantly better than the null model, which could indicate lack of model fit.
Q3: I have two groups of data, both non-normality, with bootstrap, p-value above.05. In this case, can I proceed to run multiple group CFA and text measurement invariance?
Yes, you can proceed to run multiple group CFA and testing measurement invariance. However, keep in mind the potential issues with non-normality and consider using robust methods or transformations to address these issues. Additionally, you may want to explore the results of multiple group CFA and testing measurement invariance more carefully, as the p-value above 0.05 may indicate that the model is not significantly better than the null model, which could indicate lack of fit or issues with measurement invariance.
Yuling Hu Please check if this publication helps.
https://arxiv.org/pdf/1905.05598
Best Regards
- Badges
- Science topic
- Similar topics
- Statistics
- Survey