The only significant ethnic group difference that remained was a difference in diabetes knowledge; ethnicity explained an additional 8.3% variance and emerged as a significant predictor of diabetes knowledge (β=−.308, p<.001) after controlling for SES and marital status.