Purpose Recent research suggests that the quality and type of interaction that children have with their fathers is important for children's wellbeing and development. The determinants of the nature and quality of fathers' involvement with their children are not well understood, particularly outside of the context of marriage. One important key to the father-child relationship may be mental illness, such as depression. In studying mental illness, researchers must often rely on models that employ observational data because one cannot randomly assign subjects to receive or not receive symptoms of depression. Regression analyses have been used with observational data to test the effect of depressive symptoms on parenting and child outcomes. While regression does allow for the inclusion of potentially confounding variables, it presents a number of methodological problems in making causal inferences. This poster presents a comparison of traditional regression and propensity score matching analyses measuring the influence of depression on father-child interaction and discusses some of the advantages offered by the propensity score matching techniques. Methods Ordinary Least Squares (OLS) regression analyses and propensity score matching techniques using Stata release 8.1 were employed as a secondary analysis of unique data drawn from the Fragile Families and Child Wellbeing Study (FFCWS). This longitudinal database describes the experience of a large sample of unmarried of fathers and their families. The FFCWS contains data collected on nearly 5,000 births in 20 cities across the United States using a nationally representative stratified sampling procedure of cities with populations over 200,000. A sample of 1,600 unmarried fathers is used to test the effect of depressive symptoms at birth and subsequent interaction with their children one year after birth. Results The findings in the present study suggest that a high level of depressive symptoms among unmarried fathers affects the frequency of affectionate father-child caretaking, as supported by both the regression and propensity score analyses. Using the regression analysis, fathers' depression also affects fathers' cognitive caretaking behaviors; however this relationship is not supported by the propensity score analysis. Finally, depressive symptoms do not appear to affect fathers' cognitive stimulation of their one year old children using either technique. Implications Since early parenting behaviors may set the foundation for the father-child relationship across time and unmarried fathers may be at particular risk for poor parent-child relationships, understanding the effect of paternal depressive symptoms on the unmarried father-child relationship is essential to the implementation of successful interventions with this population. Arguably some of the most interesting social work questions are not amenable to randomized control trials, either due to practical or moral restraints. Oftentimes regression analyses have been used where randomization is not possible to provide statistical control, however propensity score matching may present a strong analytic alternative. This study demonstrates that using the same survey data describing unmarried fathers and their relationship with their children, yet different statistical approaches, can result in very different conclusions about the relationship between depressive symptoms and father-child interactions. A discussion of the relative strengths and benefits of each approach is included.