Introduction
Secondary data analysis often involves a lot of considerations, particularly regarding the reliability of data and its implications. First of all, the samples used in the study may not be representative of the general population. In addition, the authors may have used inefficient data collection methods or tools. Furthermore, a lot of population surveys have missing data, especially if the surveys are filled out by the patients. These difficulties are usually accounted for in the primary articles; however, if the data is used as a source in secondary research, additional precautions have to be taken to ensure the validity and reliability of the study. One of the techniques used to analyze secondary data is weighting, which has received a wide application in secondary research (Boslaugh, 2007).
Importance of Weighting
Weighting the data is important in secondary research, as it allows to account for any misses or errors that could have occurred during the primary research process. Bell et al. (2012) state that data weighting is essential when using nationally representative health surveys. Although these surveys are large-scale and involve large numbers of respondents, data obtained from these surveys differs from the results of the studies that use simple random sampling, which creates a possibility for statistical bias (Bell et al., 2012). In most of the large-scale surveys used in secondary research, the probability of selection is unequal, as certain population groups may be left out of the surveys due to their unavailability (Bell et al., 2012). Indeed, many national surveys that are completed by patients involve respondents who are similar in terms of age and social status, whereas certain disadvantaged or affluent groups are not included. Surveys of the smaller scale may suffer from the same obstacles. Clearly, the use of weighting in secondary research in healthcare is crucial as it helps to ensure that the data is representative of the general population and not just the groups that are included in the primary study.
Uses of Weighting
Therefore, one of the widespread uses of data weighting is for the evaluation and correction of selection bias. Narduzzi, Golini, Porta, Stafoggia, and Forastiere (2013) discuss a particular technique called inverse probability weighting (IPW) for this purpose. The researchers explain that IPW was developed in the 1950s to account for the errors and bias that may result due to the non-random selection of observations. The IPW methodology “is based on the assumption that individual information that can predict the probability of inclusion (non-missingness) is available for the entire study population, so that, after taking account of them, we can make inferences about the entire target population starting from the non-missing observations alone” (Narduzzi et al., 2013, p. 335). One example of this technique in research is illustrated in a study by Van der Wal et al. (2011), where the researchers used IPW to adjust for confounding.
Another use of weighting is to account for the missing data. Seaman and White (2013) explain that restricting the data analysis to complete cases where all the needed survey results were obtained may promote bias, which is why researchers tend to avoid this method. Applying weighting to the research data that contains missing values, on the other hand, can help to remove the initial bias.
Conclusion
Overall, weighting is a valuable tool for researchers working with secondary data. The different weighting techniques can be applied to different cases in order to remove the bias, adjust the results to account for the missing values, or improve the application of the results to the general population.
References
Bell, B. A., Onwuegbuzie, A. J., Ferron, J. M., Jiao, Q. G., Hibbard, S. T., & Kromrey, J. D. (2012). Use of design effects and sample weights in complex health survey data: A review of published articles using data from 3 commonly used adolescent health surveys. American Journal of Public Health, 102(7), 1399-1405.
Boslaugh, S. (Ed.). (2007). Encyclopedia of epidemiology. Thousand Oaks, CA: Sage Publications.
Narduzzi, S., Golini, M. N., Porta, D., Stafoggia, M., & Forastiere, F. (2013). Inverse probability weighting (IPW) for evaluating and “correcting” selection bias. Epidemiologia e Prevenzione, 38(5), 335-341.
Seaman, S. R., & White, I. R. (2013). Review of inverse probability weighting for dealing with missing data. Statistical Methods in Medical Research, 22(3), 278-295.
Van der Wal, W. M., Noordzij, M., Dekker, F. W., Boeschoten, E. W., Krediet, R. T., Korevaar, J. C., & Geskus, R. B. (2011). Full loss of residual renal function causes higher mortality in dialysis patients; Findings from a marginal structural model. Nephrology Dialysis Transplantation, 26(9), 2978-2983.