Adjustment of recall errors in duration data using SIMEX

Abstract

It is widely accepted that due to memory failures retrospective survey questions tend to be prone to measurement error. However, the proportion of studies using such data that attempt to adjust for the measurement problem is shockingly low. Arguably, to a great extent this is due to both the complexity of the methods available and the need to access a subsample containing either a gold standard or replicated values. Here I suggest the implementation of a version of SIMEX capable of adjusting for the types of multiplicative measurement errors associated with memory failures in the retrospective report of durations of life-course events. SIMEX is a method relatively simple to implement and it does not require the use of replicated or validation data so long as the error process can be adequately specified. To assess the effectiveness of the method I use simulated data. I create twelve scenarios based on the combinations of three outcome models (linear, logit and Poisson) and four types of multiplicative errors (non-systematic, systematic negative, systematic positive and heteroscedastic) affecting one of the explanatory variables. I show that SIMEX can be satisfactorily implemented in each of these scenarios. Furthermore, the method can also achieve partial adjustments even in scenarios where the actual distribution and prevalence of the measurement error differs substantially from what is assumed in the adjustment, which makes it an interesting sensitivity tool in those cases where all that is known about the error process is reduced to an educated guess.

Exploring the impact of measurement error in police recorded crime rates through sensitivity analysis

“…form. Multiplicative errors are typically present in count data such as self-reported spells of unemployment (Pina-Sánchez et al, 2016), duration data such as recalled dates of developmental milestones (Pickles et al, 1996;Skinner & Humphreys, 1999), or for any other variables bound from zero to infinity that tend to be right-skewed such as personal income (Glewwe, 2007). In these instances, the measurement error term, U , is seen as proportionalrather than independent-to the unobserved true value, X .…”

The Impact of Measurement Error in Regression Models Using Police Recorded Crime Rates

“…Regardless of the complexity of the outcome model, this could be done using simulations, just as we have done in this study (the R code employed has been included in the supplementary material). Other flexible methods that could be used as sensitivity analysis tools are simulation-extrapolation (Biewen et al, 2008;Pina-Sánchez, 2016), multiple over-imputation (Blackwell et al, 2017), or Bayesian adjustments (Gustafson, 2003;Pina-Sánchez et al, 2019).…”

Exploring the Impact of Measurement Error in Police Recorded Crime Rates through Sensitivity Analysis

“…As demonstrated by Gibson and Kim (2008) and Pina-Sánchez et al (forthcoming), the measurement error affecting police recorded crime is also likely to be multiplicative in form. Multiplicative errors are typical in different forms of count data, such as self-reported spells of unemployment (Pina-Sánchez, 2016), duration data, such as in recalled dates of developmental milestones (Pickles et al, 1996;Skinner & Humphreys, 1999), or for any other variables bound from zero to infinity which tend to be right-skewed, such as estimates of income (Glewwe, 2007). In these instances, the measurement error term, U, is seen as proportional -rather than independent -to the true, unobserved, value, 𝑋.…”