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

Datos Bibliográficos

ID2330889
AutoresPaul T Von Hippel (0000-0003-4498-4374, University of Texas, Austin, TX, USA, autor de correspondencia)
Año2020
Volumen49
Número3
Páginas699-718
Fecha de publicación2020-08-01
Peer ReviewedSí
Open AccessSí
TipoARTICLE
RevistaSociological Methods & Research (JOURNAL)
Identificadores de la revistaISSN: 0049-1241 • E-ISSN: 1552-8294
EditorialSAGE Publications Inc (PUBLISHER)
DOI10.1177/0049124117747303
PMID39211325
OpenAlexW2765886208
IdiomaEN
Citas recibidas127
Referencias citadas11

When using multiple imputation, users often want to know how many imputations they need. An old answer is that 2-10 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 SAS macro called %mi_combine and a new Stata command called how_many_imputations

Data mining · Econometrics · Imputation (statistics) · Missing data · Point estimation · Quadratic equation · Standard error · Statistics · Computer Science · Mathematics · Soil Geostatistics and Mapping · Statistical Methods and Bayesian Inference · Statistical Methods and Inference