Missing information in the PMF input must be treated accordingly and excluded. Two cases can be distinguished, 1) entire row / column has no information, 2) few missing data points.
- Entire rows / columns are removed directly from the main imported input during the PMF initialization step. The process is repetitive, in that if various symbols are present for missing information, e.g. NaN and 0 for AMS data, the symbol can be changed using the corresponding button and the initialization process can be repeated. Note: the order matters and the user must start with the one that is uninterruptedly present throughout an entire row / column.
- Few missing data points are treated when defining the PMF settings. At this stage the PMF input must contain only one symbol for missing data, e.g. “NaN”, “inf”, “0” or “999”, but not more than one. Missing data that receives values that are smaller than the threshold value defined under A.II) are automatically excluded from the minimization process in ME?2. For safety reasons, their errors are also replaced by a very high value, e.g. 1000 times the highest error.
