Goodness-of-fit test for logistic regression models when data are collected using a complex sample design

Division of Epidemiology and Biostatistics and Department of Statistics, School of Public Health, The Ohio State University, 320 West Tenth Ave., M200 Starling-Loving Hall, Columbus, OH 43210, USA
Computational Statistics & Data Analysis (Impact Factor: 1.4). 05/2007; 51(9):4450-4464. DOI: 10.1016/j.csda.2006.07.006
Source: RePEc


Logistic regression models are frequently used in epidemiological studies for estimating associations that demographic, behavioral, and risk factor variables have on a dichotomous outcome, such as disease being present versus absent. After the coefficients in a logistic regression model have been estimated, goodness-of-fit of the resulting model should be examined, particularly if the purpose of the model is to estimate probabilities of event occurrences. While various goodness-of-fit tests have been proposed, the properties of these tests have been studied under the assumption that observations selected were independent and identically distributed. Increasingly, epidemiologists are using large-scale sample survey data when fitting logistic regression models, such as the National Health Interview Survey or the National Health and Nutrition Examination Survey. Unfortunately, for such situations no goodness-of-fit testing procedures have been developed or implemented in available software. To address this problem, goodness-of-fit tests for logistic regression models when data are collected using complex sampling designs are proposed. Properties of the proposed tests were examined using extensive simulation studies and results were compared to traditional goodness-of-fit tests. A Stata ado function svylogitgof for estimating the F-adjusted mean residual test after svylogit fit is available at the author's website


Available from: Kellie Archer, May 19, 2015
    • "The statistical power of the logistic regressions was determined [29]. p-Values of the Hosmer–Lemeshow test (H–L test) were also calculated [30] to describe the goodness of the models. Univariate logistic regression models were initially derived for each FE-strength and aBMD indicator in order to understand to what extent fractured and non-fractured cases could be classified using a single parameter. "
    [Show abstract] [Hide abstract]
    ABSTRACT: This is a preliminary case-control study on osteopenic/osteoporotic elderly women, testing the association of proximal femur fracture with minimum femoral strength, as derived from finite element (FE) analysis in multiple loading conditions. Fracture cases (n = 22) in acute conditions were enrolled among low-trauma fractures admitted in various hospitals in the Emilia Romagna region, Italy. Women with no history of low-trauma fractures were enrolled as controls (n = 33). Patients were imaged with DXA to obtain aBMD, and with a bilateral full femur CT scan. FE-strength was derived in stance and fall configurations: (i) as the minimum strength among those obtained for multiple loading conditions spanning a domain of plausible force directions, and (ii) as the strength associated to the most commonly used single loading conditions. The association of FE-strength and aBMD with fractures was tested with logistic regression models, deriving Odds Ratios (ORs) and the Area Under the Receiver Operating Characteristic Curve (AUC). FE-strength from multiple loading conditions better classified fracture cases from controls (OR per SD change = 9.6, 95% CI = 3.0-31.3, AUC = 0.87 in stance; OR = 9.5, 95% CI = 2.9-31.2, AUC = 0.88 in fall) compared to aBMD (OR = 3.6, 95% CI = 1.6-8.2, AUC = 0.79 for total femur aBMD), while FE-strength results from the mot commonly used single loading conditions were similar to aBMD. Only FE-strength from multiple loading conditions remained significant in age- and aBMD-adjusted models (OR = 10.5, 95% CI = 1.8-61.3, AUC = 0.95). In summary, we highlighted the importance of considering different loading conditions to identify bone weakness, and confirmed that femoral FE-strength estimates may add value to aBMD predictions in elderly osteopenic/osteoporotic women.
    Bone 10/2014; 67. DOI:10.1016/j.bone.2014.06.038 · 3.97 Impact Factor
  • Source
    • "We tested for interaction by fitting interaction terms between the main variable of interest (Aboriginal or Torres Strait Islander status) and all factors that were found to be statistically significantly associated with food insecurity. We examined the adequacy of the final model in which all factors that were associated with food insecurity were included, using the Hosmer-Lemeshow goodness-of-fit test developed specifically for complex survey data [12]. "
    [Show abstract] [Hide abstract]
    ABSTRACT: Background The prevalence of food insecurity is substantially higher among Australians of Aboriginal or Torres Strait Islander descent. The purpose of this study is to explain the relationship between food insecurity and Aboriginal and Torres Islander status in the state of Victoria. Methods Data were obtained from the 2008 Victorian Population Health Survey; a cross-sectional landline computer-assisted telephone interview survey of 34,168 randomly selected Victorians aged 18 years and older; including 339 Aboriginal and Torres Strait Islanders. We categorised a respondent as food insecure, if in the previous 12 months, they reported having run out of food and not being able to afford to buy more. We used multivariable logistic regression to adjust for age, sex, socioeconomic status (household income), lifestyle risk factors (smoking, alcohol consumption and obesity), social support (ability to get help from family, friends or neighbours), household composition (lone parent status, household with a child, and household size), and geographic location (rurality). Results Aboriginal and Torres Strait Islanders (20.3%) were more likely than their non-Aboriginal and Torres Strait Islander counterparts (5.4%) to have experienced food insecurity; odds ratio (OR) = 4.5 (95% CI; 2.7-7.4). Controlling for age, SES, smoking, obesity and inability to get help from family or friends reduced the odds ratio by 38%; ORadjusted = 2.8 (1.6-5.0). Conclusions Social determinants and lifestyle risk factors only partially explained the higher prevalence of food insecurity among Aboriginal and Torres Strait Islanders in Victoria. Further research is needed to explain the disparity in food insecurity between the two populations in order to inform and guide corrective action.
    BMC Public Health 06/2014; 14(1):598. DOI:10.1186/1471-2458-14-598 · 2.26 Impact Factor
  • Source
    • "Overall, there were very few violations of assumptions. While assessments of correct fit are not computationally available for clustered logistic regressions (Archer, Lemeshow, and Hosmer, 2007), using the Hosmer-Lemeshow test as if the data were unclustered yielded nonsignificant results for all four regressions. This might indicate correct fit. "
    [Show abstract] [Hide abstract]
    ABSTRACT: The aim of this study was to explore the role of self-beliefs in predicting postschool outcomes for deaf young adults in transition from secondary settings. Three self-level constructs were explored: self-concept, self-determination, and expectations about the future. This study utilized data from the National Longitudinal Transition Study-2 (NLTS2) that collected longitudinal data from youths with disabilities across the nation, 550 of which were deaf or hard of hearing and met the selection criteria in this study. This study examined the relationships between these deaf adolescents’ self-beliefs and actual future achievements that were reached as they transitioned to adult life, in three domains: life, employment, and education. Despite the generally positive self-beliefs of deaf individuals, which were, in some cases, related to postschool outcomes, the self-beliefs assessed in this study did not emerge as comprehensive predictors of postschool attainments. Findings suggest that for deaf individuals, successfully navigating transitions to adult life involves dimensions beyond individual agency. Positive self-beliefs are clearly a part of successfully attaining postschool outcomes, but deaf individuals may not have full access to equitable opportunities to capitalize on these beliefs.
    Journal of Developmental and Physical Disabilities 06/2014; DOI:10.1007/s10882-014-9388-y · 1.56 Impact Factor
Show more