National Forest Inventory, 2015-2023

The data analysis and editing process was carried out by IDEAM.

To carry out the validation and analysis of the data collected in the field, the IFN of Colombia has a quality assurance program, consisting of two components: (1) quality control and (2) quality evaluation.
The primary purpose of quality control is to ensure that throughout the chain of collection and custody of information, the minimum tolerance limits of each variable are met, thus ensuring the desired level of quality. To carry it out, there are manuals, guides, and formats, which have been designed to ensure that the activities are always developed, recorded and documented in the same way and that it is carried out in various stages and levels of the process: i) pre-operational field, in which training and sensitization are carried out for brigade personnel in order to ensure that the personnel know the procedures and formats in their updated versions; ii) field operation, data quality controls are applied through four levels: brigade chief, field supervisors, regional coordinators and IDEAM information criticism group. Parallel to this process, an additional control is applied, through a brigade external to the field operation team, who for each assigned operator measures seven percent (7%) of the sample through quality checks: hot, cold and blind.

Once the data has been collected, the IDEAM information control and criticism group begins the process of filtering and validating the databases, in order to correct erroneous data, through the application of algorithms, tolerance verification, criticism, validation rules, consistency and imputability. This is done both manually (through data analysts, who review forms), and automatically. For automatic data validation, a comprehensive data cleansing protocol was implemented, combining validation rules for the logical ranges that variables should have, identification of outliers, and correction by data imputation.

Imputation methods were applied, depending on the type of variable, the availability of auxiliary information and the nature of the error. In general, the following approaches were used:

• Deterministic imputation: a single value is assigned to the missing data based on a specific rule or logical procedure, without this implying randomness. This is appropriate when there is a stable and well-defined relationship between variables.
• Substitution-based imputation: values are assigned using external databases or reliable secondary sources (such as censuses, administrative records, or previous surveys). This approach corresponds to the cold cover technique.
• Regression-based imputation: statistical models (e.g., linear or logistic regression) are used to predict missing values based on other variables observed within the same record.
• Cold imputation: a representative summary value (such as mean, median, or mode) is assigned to a relevant subset of the dataset. This method is not based on individual donors, but on measures of central tendency.
• Hot Deck Imputation: a set of donor records is selected within the same dataset based on similarity to the case with missing data. A value is then assigned from this group of donors, either deterministically or randomly. This method helps preserve the internal structure and distribution of data.

STATISTICAL DISCLOSURE CONTROL (SDC)
IDEAM, as an entity that produces statistical data, is part of the National Statistical System (SEN) and as such must comply with the principles and good statistical practices defined in Law 2335 and 2023 of the Code of Good Practices of the SEN, which, according to the National Administrative Department of Statistics (DANE; 2017:11) "promotes access to and use of microdata, as well as the anonymization of microdata to ensure the protection of the identification or geographical location of sources used in the statistical process". In view of the above, and in compliance with both this standard and the NTCPE 1000:2020 Technical Standard, IDEAM applies an anonymization process to cluster coordinates through randomization.