Imagine we are working on developing a scale that measures humor style. We think that there are three orthogonal humor styles (self-deprecating, slapstick, and sarcastic) and have designed a 10-item measure that includes these uncorrelated dimensions. Items 1-5 are designed to measure self-deprecating humor, items 6-8, are designed to measure slapstick humor, and items 9 and 10 are designed to measure sarcastic humor. After running exploratory factor analysis on our set of 10 items, we realize that for the most part our set of 10 items are working well. Only item 5 seems to be a bad item, because it strongly loads on all the factors. Otherwise the items are functioning as we would expect and loading on only the factor that we expected them to load on.

Here is a correlation matrix below so that it reflects the above scenario.

	Item 1	Item 2	Item 3	Item 4	Item 5	Item 6	Item 7	Item 8	Item 9	Item 10
Item 1	-				0.9
Item 2	.9	-			0.95
Item 3	.8	.9	-		0.95
Item 4	.85	.9	.75	-	0.9
Item 5	.75	.8	.85	.8	-
Item 6	0.3	.2	.3	.2	0.9	-
Item 7	.1	.15	.3	.1	0.85	0.75	-
Item 8	.25	.1	.2	.15	0.95	.9	.8	-
Item 9	.15	.2	.25	.2	0.9	0.2	.25	.15	-
Item 10	.3	.15	.3	.15	0.95	.15	.3	.3	.95	-

1. Draw a rough scree plot we may have seen to realize there were three factors (eiganvalue scale for y doesn't need numbers)

1. Draw a plot to represent the structure of what we wanted those 10 items to look like. In other words, imagine if item 5 was actually a good item like we had hoped. Remember that ovals represent unmeasured latent factors, rectangles represent observed scores, and arrows represent relationships.