近年來，研究者對於差異試題功能（differential item functioning, DIF）議題的探討，已由「檢測」DIF轉變為「解釋」DIF。以往對於DIF試題的解釋，多有賴於專家質性審查的方式。然而，如果能有量化分析的證據輔助專家審查，可對DIF成因的判斷有所幫助。本研究透過分析DIF試題之特徵，找出試題特徵與DIF之關聯，作為後續專家審查時判斷DIF成因的參考。為此，本研究採用線性邏輯斯測驗模式（linear logistic test model, LLTM）及隨機效果線性邏輯斯測驗模式（random effects linear logistic test model, LLTM-R）針對測驗中各試題特徵進行所謂的差異層面功能（differential facet functioning, DFF）之檢測，藉以說明試題特徵與DIF之關聯。模擬研究結果顯示試題的DIF程度受到該試題特徵的DFF效果之影響。此外，測驗的Q矩陣密度較高時（例如60%），可能因型一誤差之膨脹而檢測出高比例的DIF試題；本研究另以實徵資料說明如何針對試題進行DFF分析，藉以找出與DIF有關的試題特徵，並作為後續試題修正之方向。根據結果，本研究建議採用LLTM-R進行DFF檢測，可有助於釐清試題特徵與DIF之關聯。

Because assessment methods for differential item functioning (DIF) have been developed and thoroughly investigated, the focus in DIF research has shifted to explaining DIF phenomena. Experts in this field are recruited to tap possible sources of DIF. Quantitative analysis results help experts reviewing DIF to locate sources for DIF items. This study aimed to demonstrate the use of the differential facet functioning (DFF) procedure implemented using the linear logistic test model (LLTM) and random effects linear logistic test model (LLTM-R) to explain possible DIF sources. The efficiency of LLTM and LLTM-R in detecting DFF under various conditions was also evaluated. The simulation results indicated that the DIF effect was significantly influenced by the DFF effect of item properties. Moreover, as the design matrices had a high density (e.g., 60%), Type-I error rates of DIF assessment were seriously inflated. We also demonstrated the procedure of DFF analysis with an empirical data. The result showed that most DIF items were related to two item properties, which would be provided as possible DIF sources in the item-review meeting. Researchers should implement DFF assessment using LLTM-R to help explain DIF sources.