pool.compare {mice} | R Documentation |
Compares two nested models after m repeated complete data analysis
pool.compare(fit1, fit0, data=NULL, method="Wald")
fit1 |
An object of class 'mira', produced by |
fit0 |
An object of class 'mira', produced by |
data |
In case of method "likelihood" it is necessary to pass also the original |
method |
A string describing the method to compare the two models. Two kind of comparisons are included so far: "Wald" and "likelihood". |
The function is based on the article of Meng and Rubin (1992). The Wald-method can be
found in paragraph 2.2 and the likelihoodmethod can be found in paragraph 3.
One could use the Wald method for comparison of linear models obtained with e.g. lm
(in with.mids()
).
The likelihood method should be used in case of logistic regression models obtaind with glm()
in
with.mids()
.
It is assumed that fit1 contains the larger model and the model in fit0
is fully contained in fit1
.
In case of method="Wald"
, the null hypothesis is tested that the extra parameters are all zero.
A list containing the elements:
call |
The call to the |
call11 |
The call that created |
call12 |
The call that created the imputations. |
call01 |
The call that created |
call02 |
The call that created the imputations. |
method |
The method used to compare two models: "Wald" or "likelihood" |
nmis |
The number of missing entries for each variable. |
m |
The number of imputations |
qhat1 |
A matrix, containing the estimated coeffients of the m repeated complete data analyses from |
qhat0 |
A matrix, containing the estimated coeffients of the m repeated complete data analyses from |
ubar1 |
The mean of the variances of object1, formula (3.1.3), Rubin (1987). |
ubar0 |
The mean of the variances of object0, formula (3.1.3), Rubin (1987). |
qbar1 |
The pooled estimate of object1, formula (3.1.2) Rubin (1987). |
qbar0 |
The pooled estimate of object0, formula (3.1.2) Rubin (1987). |
Dm |
The test statistic |
rm |
Relative increase in variance due to nonresponse, formula (3.1.7), Rubin (1987). |
df1 |
df1; Under the null hypothesis it is assumed that Dm has an F distribution with (df1,df2) degrees of freedom. |
df2 |
df2 |
pvalue |
P-value of testing whether the larger model is statistically different from the smaller submodel. |
Karin Groothuis-Oudshoorn and Stef van Buuren, 2009
Li, K.H., Meng, X.L., Raghunathan, T.E. and Rubin, D. B. (1991). Significance levels from repeated p-values with multiply-imputed data. Statistica Sinica, 1, 65-92.
Meng, X.L. and Rubin, D.B. (1992). Performing likelihood ratio tests with multiple-imputed data sets. Biometrika, 79, 103-111.
van Buuren S and Groothuis-Oudshoorn K (2011).
mice
: Multivariate Imputation by Chained Equations in R
.
Journal of Statistical Software, 45(3), 1-67.
http://www.jstatsoft.org/v45/i03/
### To compare two linear models: imp <- mice(nhanes2) mi1 <- with(data=imp, expr=lm(bmi~age+hyp+chl)) mi0 <- with(data=imp, expr=lm(bmi~age+hyp)) pc <- pool.compare(mi1, mi0, method="Wald") pc$spvalue # [,1] #[1,] 0.000293631 # ### Comparison of two general linear models (logistic regression). imp <- mice(boys, maxit=2) fit0 <- with(imp, glm(gen>levels(gen)[1] ~ hgt+hc,family=binomial)) fit1 <- with(imp, glm(gen>levels(gen)[1] ~ hgt+hc+reg,family=binomial)) pool.compare(fit1, fit0, method="likelihood", data=imp)