Split an untargeted metabolomics data set into a set of likely true metabolites and a set of likely measurement artifacts. This process involves comparing missing rates of pooled plasma samples and biological samples. The functions assume a fixed injection order of samples where biological samples are randomized and processed between intermittent pooled plasma samples. By comparing patterns of missing data across injection order, metabolites that appear in blocks and are likely artifacts can be separated from metabolites that seem to have random dispersion of missing data. The two main metrics used are: 1. the number of consecutive blocks of samples with present data and 2. the correlation of missing rates between biological samples and flanking pooled plasma samples.