R语言中tab_corr与apa.cor.table输出相关值差异的原因排查

Most Likely Cause: Accidental Use of Raw Data Instead of Precomputed Correlation Matrix #

The biggest culprit here is probably a typo in your apa.cor.table call: if you passed the raw data frame instead of corrmatrix, apa.cor.table uses listwise deletion by default (dropping any row with missing values in any variable). This differs from the pairwise.complete.obs method you used to compute corrmatrix, where each correlation pair uses all available data for that specific pair. If cases with missing values tend to have lower correlations, listwise deletion would exclude those rows, resulting in higher overall correlation values (0.8-0.9 range) compared to pairwise (0.5-0.7).

Other Possible Causes #

• Misreading the Output: Double-check that you’re looking at the correlation column in apa.cor.table’s output, not confidence interval bounds, p-values, or other statistics.
• Outdated Package Version: A bug in an older version of the apaTables package could cause unexpected behavior.
• Overwritten corrmatrix: Ensure the corrmatrix object in your environment hasn’t been overwritten by another variable with the same name (though this is less likely since tab_corr matches its values).

Actionable Advice #

• Verify Your Code: Confirm you’re running apa.cor.table(corrmatrix) and not apa.cor.table(data).
• Inspect the Correlation Matrix: Run print(corrmatrix) to explicitly check that its values are indeed in the 0.5-0.7 range.
• Update apaTables: Install the latest version with:

  install.packages("apaTables")
  

• Use Pairwise Deletion with Raw Data (If Needed): If you want to compute correlations directly from data using pairwise deletion in apa.cor.table, specify the na.action argument:

  apa.cor.table(data, na.action = na.pairwise)
  

  This will match the correlation calculation method used for your corrmatrix.

内容的提问来源于stack exchange，提问作者Laura Ashlock