Questionnaire methodology and user testing
Questionnaire methodology includes a number of different methods that are used iteratively. The field of study is based on a number of other academic disciplines such as social anthropology, psychology, linguistics and philosophy. The cognitive perspective on the response process is central to our work, and can be divided into four iterative phases: interpretation of meaning, retrieval of information required, assessment of information, and the selection of an answer. Our methods for user testing include: expert evaluation, focus groups, cognitive mapping, eye tracking and usability testing.
See more in Practical user testing
A sample survey is conducted when it is too extensive to ask the entire population. Sampling design is the planning of which units are to be included in such a survey. This includes analyzing the population that is the basis for the sample, calculating sample sizes and determining the sampling method.
Data integration is to combine data from different data sources, with the aim of producing new data sets that are the basis for statistics or research.
Data editing is the control, scrutiny and correction of data. It includes editing of population, editing of obvious and systematic errors, selection of values with large deviations and high influence and control of aggregates to be published. The methods used for data editing range from logical control of valid value range to imputation of values with machine learning.
See more in Generic Statistical Data Editing Model GSDEM
Estimation and weighting
Estimation is to find value for a population figure based on the information we have collected from the (sample) survey. We are usually interested in several figures - totals, averages, proportions and variances are most common - for several variables. Estimation often means that each unit in the sample is assigned a weight, this is almost always done for personal and household surveys. We can also base the estimation on a statistical model, which is common in business surveys.
Seasonal adjustment is to use statistical methods to remove systematic seasonal variations from a monthly or quarterly time series, so that the time series to the greatest possible extent expresses the real development over time. In addition, they try to remove the calendar effects that vary from year to year, such as Easter. When the data has been corrected for the seasonal conditions, one will be left with a clearer picture of the underlying development in the time series which consists of trend / cycle and irregular component.
The Statistics Act § 7. Statistical confidentiality in the dissemination of official statistics requires that Statistics Norway does not publish statistics in a manner that allows statistical information to be traced back to individuals or other types of statistical units. Confidentiality can be ensured by using coarser categories, by suppression (hiding values) or by perturbation (changing values). Examples of coarser categories are counties instead of municipalities and age groups of five years instead of single years. Suppression means that some values are not published and are instead replaced by a colon. In addition to the values one wishes to protect via suppression, several other values in the table must be so-called secondary suppressed to prevent the possibility of recalculation from other published values (aggregates). When applying perturbation, the published value may deviate from the real value. Several methods can be used, including rounding and noise addition.