What Improves with Increased Missing Data Imputations?

Bodner, Todd

doi:10.1080/10705510802339072

Cited by 503 publications

(344 citation statements)

References 11 publications

Supporting

Mentioning

316

Contrasting

Unclassified

Order By: Relevance

“…Figure 1 fits Bodner's results (2008, Table 3, column 2) to our quadratic rule which, with a CV of .05, simplifies to 200 . Figure 1 also shows the linear rule 100 proposed by others (Bodner, 2008;White et al, 2011). Clearly the quadratic rule fits better.…”

Section: Why the Estimated Df Is An Unreliable Guidementioning

confidence: 97%

“…But this linear rule understates the required number of imputations when is large (Bodner, 2008) and overstates the required number of imputations when is small (see Figure 1). …”

Section: A Quadratic Rulementioning

confidence: 99%

“…Until recently, it was standard advice that M = 2 to 10 imputations usually suffice (Rubin, 1987), but that advice only addresses the relative efficiency of the point estimate (see Section 4.1). If you also want the SE estimate to be replicable, then you may need more imputations-as many as M > 200 in some situations-though fewer imputations may still suffice in many settings (Bodner, 2008;Graham, Olchowski, & Gilreath, 2007;White, Royston, & Wood, 2011), It might seem conservative to set M = 200 (say) as a blanket policy-and that policy may be practical in situations where the data are small and the imputations and analyses run quickly. But if the data are large, or the imputations or analyses run slowly, then you will want to use as few imputations as you can, and you need guidance regarding how many imputations are really needed for their specific data and analysis.…”

Section: Overviewmentioning

confidence: 99%

“…A recently proposed rule is that, to estimate SEs reliably, you should choose the number of imputations according to the linear rule 100 (Bodner, 2008;White et al, 2011), where is the fraction of missing information, defined below in equation (4). But this linear rule understates the required number of imputations when is large (Bodner, 2008) and overstates the required number of imputations when is small (see Figure 1).…”

Section: A Quadratic Rulementioning

confidence: 99%

See 3 more Smart Citations

How Many Imputations Do You Need? A Two-stage Calculation Using a Quadratic Rule

Hippel

Paul²

2018

Sociological Methods & Research

312

200

View full text Add to dashboard Cite

Abstract0 FWhen using multiple imputation, users often want to know how many imputations they need. An old answer is that 2 to 10 imputations usually suffice, but this recommendation only addresses the efficiency of point estimates. You may need more imputations if, in addition to efficient point estimates, you also want standard error (SE) estimates that would not change (much) if you imputed the data again. For replicable SE estimates, the required number of imputations increases quadratically with the fraction of missing information (not linearly, as previous studies have suggested). I recommend a two-stage procedure in which you conduct a pilot analysis using a small-to-moderate number of imputations, then use the results to calculate the number of imputations that are needed for a final analysis whose SE estimates will have the desired level of replicability. I implement the two-stage procedure using a new Stata command called how_many_imputations (available from SSC) and a new SAS macro called %mi_combine (available from the website missingdata.org).

show abstract

Section: Why the Estimated Df Is An Unreliable Guidementioning

confidence: 97%

“…But this linear rule understates the required number of imputations when is large (Bodner, 2008) and overstates the required number of imputations when is small (see Figure 1). …”

Section: A Quadratic Rulementioning

confidence: 99%

Section: Overviewmentioning

confidence: 99%

Section: A Quadratic Rulementioning

confidence: 99%

See 2 more Smart Citations

How Many Imputations Do You Need? A Two-stage Calculation Using a Quadratic Rule

Hippel

Paul²

2018

Sociological Methods & Research

312

200

View full text Add to dashboard Cite

show abstract

“…This variation captures the estimation uncertainty due to missingness, which is called the between-imputation variance (Little & Rubin, 2002). Obviously, real applications require a much larger value of M (Graham et al, 2007;Bodner, 2008).…”

Section: Em Algorithmmentioning

confidence: 99%

Multiple ratio imputation by the EMB algorithm: theory and simulation

Takahashi

2017

J. Mod. App. Stat. Meth.

View full text Add to dashboard Cite

show abstract

Dupilumab Drug Survival and Associated Predictors in Patients With Moderate to Severe Atopic Dermatitis

et al. 2022

View full text Add to dashboard Cite

ImportanceLong-term data on dupilumab drug survival in patients with atopic dermatitis (AD) are scarce. Furthermore, little is known about the factors associated with drug survival of dupilumab in AD.ObjectiveTo describe the drug survival of dupilumab in patients with AD and to identify associated predictors.Design, Setting, and ParticipantsThis cohort study was based on data from the multicenter prospective daily practice BioDay registry, in which 4 university and 10 nonuniversity hospitals in the Netherlands participated. Analysis included patients (age ≥18 years) participating in the BioDay registry with a follow-up of at least 4 weeks. The first patient treated with dupilumab was recorded in the BioDay registry in October 2017; data lock took place in December 2020, and data analysis was performed from October 2017 to December 2020.Main Outcomes and MeasuresDrug survival was analyzed by Kaplan-Meier survival curves and associated characteristics by using univariate and multivariate Cox regression analysis.ResultsA total of 715 adult patients with AD (mean [SD] age, 41.8 [16.0] years; 418 [58.5%] were male) were included with a 1-year, 2-year, and 3-year overall dupilumab drug survival of 90.3%, 85.9%, and 78.6%, respectively. Characteristics associated with shorter drug survival owing to ineffectiveness were the use of immunosuppressant drugs at baseline (hazard ratio [HR], 2.64; 95% CI, 1.10-6.37) and being a nonresponder at 4 weeks (HR, 8.68; 95% CI, 2.97-25.35). Characteristics associated with shorter drug survival owing to adverse effects were the use of immunosuppressant drugs at baseline (HR, 2.69; 95% CI, 1.32-5.48), age 65 years or older (HR, 2.94; 95% CI, 1.10-7.87), and Investigator Global Assessment score of very severe AD (HR, 3.51; 95% CI, 1.20-10.28).Conclusions and RelevanceThis cohort study demonstrated a good overall 1-year, 2-year, and 3-year dupilumab drug survival. Patients using immunosuppressive therapy at baseline and those with an absence of treatment effect at week 4 tended to discontinue treatment owing to ineffectiveness more frequently. Using immunosuppressant drugs at baseline, older age, and Investigator Global Assessment score of very severe AD were characteristics associated with an increased risk for discontinuation owing to adverse effects. These data provide more insight and new perspectives regarding dupilumab treatment in AD and can contribute to the optimization of patient outcomes.

show abstract

What Improves with Increased Missing Data Imputations?

Cited by 503 publications

References 11 publications

How Many Imputations Do You Need? A Two-stage Calculation Using a Quadratic Rule

How Many Imputations Do You Need? A Two-stage Calculation Using a Quadratic Rule

Multiple ratio imputation by the EMB algorithm: theory and simulation

Dupilumab Drug Survival and Associated Predictors in Patients With Moderate to Severe Atopic Dermatitis

Contact Info

Product

Resources

About