EUSN 2026, Norrkøping


Advanced RSiena workshop WS21

August 15, 2026

8.30 AM - 12.00 AM, 14.00 PM - 17:30 PM
Room TP45



Programme

  1.   Transition to the new function names

  2. Parameter interpretation: semi-standardized parameters: entropy-based approach to explained variation.

  3.   Testing; Score-type tests .


  4.   Problems with convergence: various kinds .

  5.   Interactions, elementary effects, contextual effects

  6.   Multivariate networks.

  7. Two-mode networks, and their co-evolution with one-mode networks

  8.   Analysis of multilevel networks .

  9.   Missing data .

    • Slide presentations:

    • Scripts :

    • Literature :

      • Huisman, Mark, and Steglich, C.E.G., (2008). Treatment of non-response in longitudinal network data.
        Social Networks, 30, 297-308.
      • Robert W. Krause, Mark Huisman, and Tom A.B. Snijders (2018). Multiple imputation for longitudinal network data. Italian Journal of Applied Statistics, 30, 33-57.
        DOI: https://doi.org/10.26398/IJAS.0030-002.

        Missing data on network ties are a fundamental problem for network analysis. In this paper, we present a new method with two variants to handle missing data due to actor non-response in the framework of stochastic actor-oriented models (SAOMs). The proposed method imputes missing tie variables in the first wave either by using a Bayesian exponential random graph model (BERGMs) or a stationary SAOM and imputes missing tie variables in later waves utilizing a SAOM. The proposed method is compared to the standard SAOM missing data treatment as well as recently proposed methods. The multiple imputation procedure provided more reliable point estimates than the default treatment.

      • Kayla de la Haye, Joshua Embree, Marc Punkay, Dorothy L. Espelage, Joan S. Tucker, and Harold D. Green Jr. (2017). Analytic strategies for longitudinal networks with missing data. Social Networks, 50, 17-25.
        DOI: https://dx.doi.org/10.1016/j.socnet.2017.02.001
        Proposes a strategy for dealing with varying actors subsets having missing data. The strategy is based on the multiple groups option.
      • Hipp, J.R., Wang, C., Butts, C.T., Jose, R., Lakon, C.M. (2015). Research note: the consequences of different methods for handling missing network data in stochastic actor based models. Social Networks 41, 56-71.
        DOI: https://dx.doi.org/10.1016/j.socnet.2014.12.004.

        Note
        This paper presents a method for handling missing data in longitudinal network modeling that is better than the simple method employed in RSiena. However, the method in RSiena is not as poor as these authors suggest. The paper misrepresents details of the way in which RSiena handles missing data. The writing suggests that the 'third strategy' mentioned on p. 57 is the strategy employed by RSiena. This is not the case, because although RSiena uses the imputation strategy as described here, it also excludes the tie and behavior variables for which a data values is missing at a wave and/or a preceding wave from the calculation of the target statistics for this wave. Thus, the effect of the imputation is minimized. See Huisman and Steglich (2008, Section 4.4), and the RSiena Manual, Section 5.3.2.

      • Cheng Wang, Carter T. Butts, John R. Hipp, Rupa Jose, and Cynthia M. Lakon (2016). Multiple imputation for missing edge data: A predictive evaluation method with application to Add Health. Social Networks 45, 89-98.
        DOI: https://dx.doi.org/10.1016/j.socnet.2015.12.003.

        Note
        See the note about the paper by Hipp et al. (2015), a few lines up. This paper likewise misrepresents the way in which RSiena handles missing data. The phrase 'it simply drops them or treats them as absent' (p. 95) is incorrect: the imputation used in RSiena by 'last observation carried forward' is somewhat more informative than this, and the exclusion of the imputed data from the target statistics means that 'simply' is not applicable. Further, 'As Hipp et al. (2015) found' is incorrect, because Hipp et al. (2015) used a different imputation strategy in their comparisons, as mentioned above.

      • Tjeerd Zandberg and Mark Huisman (2019). Missing behavior data in longitudinal network studies: The impact of treatment methods on estimated effect parameters in stochastic actor oriented models. Social Network Analysis and Mining, 9.8.
        DOI: https://doi.org/10.1007/s13278-019-0553-2.

        This paper examines seven different methods that are currently available to deal with missing behavior data. The effect of the missing data methods was inspected using three criteria: model convergence, parameter bias, and parameter coverage. The results show that, in general, the default method available in the RSIENA software gives the best outcomes for all three criteria.

  10.   Some effects that are little known, but which may be useful for analyzing two-mode networks.

  11. If time permits:

    Non-directed networks.

  12. If time permits:

    Valued networks (two kinds: networks with weak and strong ties; signed networks).

  13. If time permits:

      Internal effect parameters and interaction effects

  14.   And....




Back to the main Siena page