‡Missing values: Belgium no missing values, the Netherlands N = 15, Italy N = 134, Spain N = 6.
*Missing values: Belgium no missing values, the Netherlands N = 2, Italy no missing values, Spain N = 2.
Fig. 3 The first plot for estimates of missing values; the second plot for the errors in estimating missing values.
Fig. 4 The first plot is for estimates of missing values; the second plot is for the errors in estimating missing values.
In questionnaires with less than 20% missing values, the method of 'case mean substitution' was used to replace the missing value.
After deleting respondents with missing values, the sample size reduced to n = 2851 observations.
Simulations show that for arbitrary settings of complete data and missing values, the resulting confidence intervals maintain the pre-assigned coverage probability quite accurately.
In the case of missing values the patients were interviewed in detail about frequency, duration and intensity of their headaches through the preceding four weeks.
Under an MAR condition with a maximum of 20% of missing values, the differences in group means were estimated without bias.
To better understand the strengths and limitations of the data, it also includes checking data completeness, redundancy, missing values, the plausibility of attribute values.
Based on the pattern of the missing values, the proposed approach finds subsets of samples for which most of the features are available and trains a classifier for each subset.
