UBC Theses and Dissertations
A psychometric analysis of the Hooked on Nicotine Checklist (HONC) with an eye towards gender differential item functioning : a case study in missing data and DIF Zeisser, Cornelia
This study investigated gender Differential Item Functioning (DIF) in a measure of nicotine dependence (ND) in adolescents, the Hooked on Nicotine Checklist (HONC). First, a statistical modeling technique based on binary logistic regression was used to determine the presence and extent of DIF for each HONC item. Second, graphical DIF analyses were performed using nonparametric item response theory (NIRT). To investigate the impact of missing data on findings of DIF, all DIF analyses were performed on four different versions of the data: a) listwise deletion of missing cases (no imputation), b) imputation of missing values by row mode, c) imputation of missing values by column mode and d) imputation of missing values using the expectation maximization (EM) algorithm based on maximum likelihood estimation. Using binary logistic regression analysis, none of the ten HONC items were flagged as displaying DIF, with identical results across all four versions of the data. Using NIRT for graphical displays of DIF, seven out of ten HONC items showed DIF under column-wise and EM imputation of missing values, while eight out of ten HONC items were flagged as DIF under no imputation and row-wise imputation of missing values. The study concluded that missing data techniques did not have a strong influence on finding DIF. However, the importance of conducting DIF analyses with various DIF methods, including graphical displays, is emphasized, as the items displayed no DIF under logistic regression analyses, but marginal to substantial DIF using the NIRT. To potentially improve the HONC, it appears worth considering other possible dimensions to be included in the HONC as a measure of adolescent tobacco dependence pertaining to the psychological-social aspects of tobacco dependence in this particular population.
Item Citations and Data