What emerges when you extract 2 factors? 5? 14? Explore how the structure of belief unfolds as you increase the number of factors.
*All summary names are generated and may not be accurate.
Dataset. ~32,000 respondents answered up to 1,084 agree/disagree items on a 7-point Likert scale (strongly disagree → strongly agree). After cleaning, 23 items were removed (redundant duplicates and surveyor-specific items), leaving 1,061 items. All 1,061 items are included in the analysis — no pre-filtering.
Factor extraction. For each solution (2–18 factors), PCA extracts components from all items, followed by oblimin (oblique) rotation. Oblique rotation allows factors to correlate, which better reflects how real attitudes cluster. Each item is assigned to the factor on which it loads most strongly (≥ |0.25|).
Subfactors. Factors with 10+ items are tested for internal structure. Parallel analysis determines whether multiple subfactors are warranted. If so, a nested PCA + oblimin rotation extracts them, with low-quality subfactors discarded.
What the numbers mean. A loading indicates how strongly an item relates to a factor — higher magnitude means a stronger relationship. A positive loading means agreement with the item tracks with that pole of the factor; a minus sign means agreement tracks with the opposite pole. Variance explained is the share of total item variation captured by the extracted factors.
Interpreting the sweep. At 2 factors you see the broadest ideological split. As factor count increases, large factors subdivide into more specific dimensions. There is no single "correct" number — different levels reveal different structure.