Cambodia Agriculture Survey 2024

Once the enumerators collected the survey data for an agricultural household, they submitted the completed questionnaire via Survey Solutions to their data supervisors who, in turn, carried out scrutiny checks. If there were errors or suspicious data detected, the data supervisor would return the record to the enumerator to address the issues with the respondent if needed, and the corrected record would be re-submitted to the data Supervisor. Once the records were validated by the data supervisors, they would approve them for final review by headquarters staff.

At the survey headquarters, the completed questionnaires were received after being approved by the data supervisors. If any issues or suspicious data were discovered during the headquarters review, the records could be returned to the enumerator for verification or correction if needed. Documentation on how to review questionnaire data for suspicious items or outliers was provided to both data Supervisors and headquarters staff.

The data review and calculation of the survey estimates was undertaken using the RStudio software tool. Validation of the data began even when the questionnaires were being designed in the CAPI tool, as Survey Solutions allows for consistency checks to be built into the data collection tool. As soon as completed records were returned during the data collection stage, additional consistency checks were completed, evaluating the ranges for certain items, and verifying any outlier records with the enumerator and/or respondent. Moreover, when the data was cleaned, another step was conducted to impute the missing values derived from item non-response.

STATISTICAL DISCLOSURE CONTROL (SDC)

Microdata are disseminated as Public Use Files under the terms and conditions indicated at the NIS Microdata Catalog (https://microdata.nis.gov.kh/), as indicated in the section about 'access conditions' below.

In addition, anonymization methods have been applied to the microdata files before their dissemination, to protect the confidentiality of the statistical units (e.g. individuals) from which the data were collected. These methods include: i) removal of some variables contained in the survey (e.g. name, address, etc.), ii) grouping values of some variables into categories (e.g. age categories), iii) limiting geographical information to the province level, iv) removal of some records or specific data points, v) censoring the highest values in continuous variables (top-coding) by groups, replacing them with less extreme values from other respondents, or vi) rounding numerical values.

Also, some variables were recoded to group some options together and therefore lower the risk of identifying specific respondents with unusual characteristics, similar to previous years. One example of such is the crop roster, where the main 34 crops (the same as in the CAS2024 reports and Camstat) have been maintained separate, but the rest have been recoded into broader categories. This may however, result in some duplicated values as a result of this recoding for the broader categories. Same applies to any roster with crop information such as the crop seed roster and crop harvested roster. Ultimately these recoded categories are other crop codes with lower analytical values, compared to the main 34 crops.

Users must therefore be aware that data protection with SDC methods involves perturbations in the microdata. This implies information loss and bias, and affects the resulting estimates and their parameters. In general, the smaller the subpopulation, the higher the potential impact derived from the anonymization process.