In September of 2014, we conducted coordinated data collections using an ABS probability mail survey (Westat), a probability web survey (Pew Research Center), and a non-probability web survey using SurveyMonkey’s Audience panel (SurveyMonkey). A separate paper compares weighted estimates from these surveys, including comparisons to national benchmark estimates. This paper examines two statistical approaches that have been proposed to improve the quality of the estimates from non-probability samples. The first method involves using different weighting. Most weighting strategies have examined approaches intended to address coverage loss associated with lack of access or use of the internet. Propensity score weighting is one such approach. We explore alternative calibration weights, informed by data from the probability-based Westat and Pew surveys. The second approach is sample matching, where the respondents for the non- probability web survey are selected to match to targets from other samples or censuses. Since sample matching was not implemented in the SurveyMonkey web survey, we simulate this option by dropping observations from the data set so that the unweighted sample is more similar to observations from the probability mail survey. We explore the effects of both of these techniques by comparing the web survey estimates with and without the weighting and matching adjustments to the other probability sample estimates and benchmarks discussed above.