Coates Jennifer, Wilde Parke E, Webb Patrick, Rogers Beatrice Lorge, Houser Robert F
Gerald J. and Dorothy R. Friedman School of Nutrition Science and Policy, Tufts University, Boston, MA 02111, USA.
J Nutr. 2006 May;136(5):1420S-1430S. doi: 10.1093/jn/136.5.1420S.
This paper compares a qualitative and a quantitative (Rasch) method of item assessment for developing the content of a food insecurity scale for Bangladesh. Data are derived from the Bangladesh Food Insecurity Measurement and Validation Study, in which researchers collected 2 rounds of ethnographic information and 3 rounds of conventional household survey data between 2001 and 2003. The qualitative method of scale development relied on content experts and respondents themselves to evaluate household food insecurity items generated through ethnographic research. The quantitative method applied the Rasch model to assess the fit of the same items using representative survey data. The Rasch model was then used to test for differential item functioning (DIF) across diverse demographic and geographic subgroups. The qualitative assessment flagged and discarded 10 items, leaving 13. The Rasch assessment of infit and outfit flagged 3 items, and the Rasch DIF test discarded another 10 items, leaving a total of 10 items in the Rasch-derived scale. The 2 scales contained 8 of the same items. The qualitatively and quantitatively derived scales were highly correlated (r = 0.96, P < 0.01), and the 2 methods located 90% of households in the same food insecurity tercile. This convergence lends added confidence to the use of either scale for identifying food-insecure households in different regions of Bangladesh. Multiple methods should continue to be applied in a systematic and transparent way to lend additional credence to the results when they converge and to pinpoint directions for further clarification where they do not.
本文比较了用于制定孟加拉国粮食不安全量表内容的定性和定量（拉施）项目评估方法。数据源自孟加拉国粮食不安全测量与验证研究，在该研究中，研究人员于2001年至2003年间收集了两轮人种志信息和三轮传统家庭调查数据。量表开发的定性方法依靠内容专家和受访者自身来评估通过人种志研究生成的家庭粮食不安全项目。定量方法应用拉施模型，使用代表性调查数据评估相同项目的拟合度。然后使用拉施模型测试不同人口和地理亚组间的项目功能差异（DIF）。定性评估标记并剔除了10个项目，剩下13个。拉施模型对内在拟合和外在拟合的评估标记出3个项目，拉施DIF测试又剔除了10个项目，拉施衍生量表中总共剩下10个项目。两个量表包含8个相同项目。定性和定量得出的量表高度相关（r = 0.96，P < 0.01），两种方法将90%的家庭划分到相同的粮食不安全三分位数中。这种趋同为在孟加拉国不同地区使用任一量表识别粮食不安全家庭增添了信心。应继续以系统且透明的方式应用多种方法，以便在结果趋同时为其增添可信度，并在结果不同时明确进一步澄清的方向。
