Skip to main content



User Support

NIS-4 User Support

Before beginning NIS-4 analyses, read the READ_ME_FIRST.pdf included in your dataset and review the NIS-4 PUF Manual (PDF) and CodebooK (HTML).

NIS-4 uses a complex sample design and paired jackknife replicate weighting (JK2). In replicate weighting, subsamples (replicates) are drawn from the sample and the statistic of interest is calculated for each subsample. Then, the variability among these subsamples is used to estimate the variance of the full sample statistic. For more information on replicate weights for variance estimation in NIS-4, see Chapter 5 in the NIS4_Analysis_Report available from www.nis4.org and included in the dataset.

You MUST use the weights when working with NIS-4 data.
You are encouraged to calculate known estimates from the NIS-4 Report to Congress to confirm correct use of the replicate weights.

Statistical software, including SPSS, can be used for data manipulation, or to investigate totals and rates (without computing confidence intervals, fitting models, investigating relationships among variables, or conducting comparisons.)

Because of the complex sample design and use of paired jackknife replicate weighting (JK2), users have been advised in the past to conduct analyses only with WesVar or SUDAAN (in particular, for significance tests or compute any statistic that relies on variance or standard error). Preliminary analyses (frequencies and logistic regression) demonstrate that you can analyze NIS-4 using SAS, Stata, R, WesVar, WesDaX, or SUDAAN software. (NDACAN does not provide support for SUDAAN).

The results among software packages may differ slightly due to differences in how the packages calculate standard error: some use the estimate based on the full data in the calculation of the standard error, whereas others use the average estimate over the replicate weights. There is no clear consensus on which is better. The difference is negligible in tested cases.

Documents

Videos and Presentations


©2017 College of Human Ecology, Cornell University. All rights reserved.