Treatment of Missing Data

List-wise deletion would have the effect that the three variables mentioned above result in a loss of 30.3 % of all cases. 

For the statistical questions of interest here, it was useful to carry out two preliminary steps of data preparation imputation and augmentation. 

In the augmentation step, secondary data sources describing the vehicle profile were fused with the GIDAS data set, as the coded information is very sparse, as not many variables describing vehicle profile characteristics are included in GIDAS. Those characteristics are supposed to influence injury severity and thus should be 

Important geometric quantities for the PCDS data set are given Figs. A.4, p. 181, and A.5, p. 181. Corresponding measurements for the pedestrian are illustrated in Fig. A.6, p. 182, both for GIDAS and PCDS. 

Impact speed was not imputed in either data set as it is the most important predictor for injury causation as mentioned above. 

---

Howell DC (2002) Treatment of missing data. www.uvm.edu/ dhowell/StatPages/ More Stuff/Missing Data/Missing.html... 

The report should contain a clear statement of procedures for dealing with the data and the determination of test statistics (e.g., treatment of missing data, outhers, and method of calculation of z-scores). Usually quoted are the valid measurement results. 

O Neill RT, Temple R (2012) The prevention and treatment of missing data in clinical trials an FDA perspective on the importance of dealing with it. Clin Pharmacol Ther 91 550-554... 

A similar treatment of the data of Ojima et al. shows that 0.66 mmol each of Me and Si and 3.3 mol of H are unaccounted for by the yields of products given in Eq. (32). These missing quantities correspond to a 26% yield each of MeSiH3 and H2. The higher yield of H2 in the Rh-cata-lyzed reaction vis-a-vis the Ir-catalyzed reaction is consistent with the formation of di- and trisilanes with the Rh catalyst. For each mole of Si—Si bonds formed, 1 mol of H2 equivalent must be evolved [Eq. (35)]. The H2 may enter into the general redistribution scheme before escaping from solution and hence may not appear exclusively as H2, but may also contribute to the hydrogen-rich silane fraction. 

Hot deck is another type of single imputation. This procedure matches individuals with missing data with those having similar values in a set of other variables and imputes the known value into the missing data cell. The flaw with hot deck imputation is its treatment of imputed data with certainty, thus perhaps grossly underestimating variability (6). 

Although the issues of missing data and intention to treat are closely related, they are not identical. Intention to treat analysis is possible provided outcome and randomization data are not missing and (in its strict sense) impossible if they are. If data are not missing, then even if patients are noncompliant it is possible to analyse them as randomized provided only that their original treatment allocation is known. A per-protocol analysis requires knowledge of compliance. Also, depending on what one means by per protocol, it may or may not require complete outcome data. For example, one could simply say for a per-protocol analysis that one is only interested in patients who not only took the medicine as instructed but also recorded their outcomes. 

As it has been discussed in the literature, we also compare the performance of straight Fourier transformation FFT on the NUS synthetic four-line spectrum in Fig. 3. We used the schedules obtained with the best seed numbers as in Fig. 2a and leave the missing time domain data points at zero value. The traces presented in Fig. 3 show that the artifacts due to the sampling schedules and lack of reconstruction are severe and mask the small peaks. However, the intense line is readily observable. Thus, straight Fourier transformation may be an option if one is only concerned with very intense peaks, such as methyl resonances in a protein, and weak peaks are of little concern. This treatment of NUS data may be useful for a quick inspection of an NUS data set to find out whether an experiment has worked. However, it should be followed by a reconstruction effort to retrieve best the full information content of the NUS data. 

One issue that is important to determine is how missing data will be handled in an ITT analysis. ITT was largely developed in clinical trials in which the major endpoints were events, mortality, infarctions, etc. In such studies it is possible to follow-up on patients who withdraw from treatment and determine whether the event has occurred or not and indeed in such trials every effort should be made to do this. For other types of data, for example, a pain scale other approaches are necessary. 

One very simplistic way of handling missing data is to remove those patients with missing data from the analysis in a complete cases analysis or completers analysis. By definition this will be a per-protocol analysis which will omit all patients who do not provide a measure on the primary endpoint and will of course be subject to bias. Such an analysis may well be acceptable in an exploratory setting where we may be looking to get some idea of the treatment effect if every subject were to follow the protocol perfectly, but it would not be acceptable in a confirmatory setting as a primary analysis. 

On receipt of the measurement results, the organizing laboratory collates the data and statistically analyzes the results. The treatment of outliers and missing data is important in any type of study. 

---