Large-sample significance levels from multiply imputed data using moment-based statistics and an F reference distribution

摘要: Abstract We present a procedure for computing significance levels from data sets whose missing values have been multiply imputed data. This uses moment-based statistics, m ≤ 3 repeated imputations, and an F reference distribution. When = ∞, we show first that our is essentially the same as ideal in cases of practical importance and, second, its deviations are basically function coefficient variation canonical ratios complete to observed information. For small procedure's performance largely governed by this mean these ratios. Using simulation techniques with m, compare actual nominal large-sample conclude it calibrated thus represents definite improvement over previously available procedures. Furthermore, power other factors, such di...