Skip to contents

Lists the items, or for validity the constructs, whose statistics cross the cut-off usually applied to them, with the value, the rule, its source and a ready reason, such as "org_4: loading .21 in RQ3, below .40". A flag is a prompt to review the item, not a rule to drop it: its content, wording, reverse keying and the purpose of the model still decide. Drop an item with sframe_drop_items() and record it with amend_sframe().

Usage

sframe_item_flags(result, htmt = 0.85, all = FALSE, block = NULL)

Arguments

result

A result of one of the kinds above.

htmt

The HTMT threshold: 0.85 for conceptually distinct constructs (the default) or 0.90 for similar ones. The value applied is kept as the attribute htmt_threshold.

all

If TRUE, every row is returned, flagged or not.

block

Character or NULL. The plan block named in the reasons. A plan result supplies its own.

Value

A data frame with columns result_type, block, group (the scale, factor, construct or construct pair), item, statistic, value, cutoff, direction, flagged, reason and source.

Details

The results read, and the rule applied to each:

  • a reliability result (sframe_reliability_results()): alpha if the item is deleted above the scale's alpha. Alpha can rise when a scale narrows, so the reason asks for the item-total correlation and the wording to be checked too.

  • an item result (sframe_item_results()): corrected item-total correlation below .30 (Nunnally and Bernstein, 1994).

  • an EFA readiness result (sframe_efa_readiness_results()): item sampling adequacy (MSA) below .50 (Kaiser, 1974).

  • an EFA solution (sframe_efa_solution()): primary loading below .40 (Hair et al., 2019), and, with 2 or more factors, the 2 largest loadings less than .20 apart (Howard, 2016), as 2 separate reasons.

  • a CFA fitted with lavaan, as from the syntax sframe_syntax_cfa() writes: standardised loading below .50 (Hair et al., 2019).

  • a validity result (sframe_validity_results()): AVE below .50 (Fornell and Larcker, 1981), and HTMT above htmt (Henseler et al., 2015).

  • a PLS model estimated with seminr: outer loading on the item's own construct below .708 (Hair et al., 2022). Below .40 the item is a strong removal candidate. From .40 to .708, check AVE and composite reliability before removing it.

A result from run_analysis_plan() for one of these methods is read the same way, and its reasons name the block.

References

Fornell, C., & Larcker, D. F. (1981). Evaluating structural equation models with unobservable variables and measurement error. Journal of Marketing Research, 18(1), 39-50. doi:10.1177/002224378101800104

Hair, J. F., Black, W. C., Babin, B. J., & Anderson, R. E. (2019). Multivariate data analysis (8th ed.). Cengage.

Hair, J. F., Hult, G. T. M., Ringle, C. M., & Sarstedt, M. (2022). A primer on partial least squares structural equation modeling (PLS-SEM) (3rd ed.). Sage.

Henseler, J., Ringle, C. M., & Sarstedt, M. (2015). A new criterion for assessing discriminant validity in variance-based structural equation modeling. Journal of the Academy of Marketing Science, 43(1), 115-135. doi:10.1007/s11747-014-0403-8

Howard, M. C. (2016). A review of exploratory factor analysis decisions and overview of current practices. Journal of Business Research, 69(1), 51-62. doi:10.1016/j.jbusres.2015.08.024

Kaiser, H. F. (1974). An index of factorial simplicity. Psychometrika, 39(1), 31-36. doi:10.1007/BF02291575

Nunnally, J. C., & Bernstein, I. H. (1994). Psychometric theory (3rd ed.). McGraw-Hill.

Revelle, W. (2025). psych: Procedures for psychological, psychometric, and personality research (R package). Northwestern University. https://CRAN.R-project.org/package=psych

Examples

d <- sframe_demo("likert_scale")
if (requireNamespace("psych", quietly = TRUE)) {
  sframe_item_flags(sframe_item_results(d$responses, d$instrument), all = TRUE)
}