Information on specific statistical methods

This page provides information about specific methods used in the process to produce statistics across the Government Statistical Service (GSS).

For each of these methods there is an expert group assigned within the Methodology division at the Office for National Statistics (ONS).  We also have expert members from across the GSS in the groups. The aim of the Expert Groups is to maintain and share best practice and build expertise and capability in their specific areas.  These groups are open to experts across the GSS and academia.

Each section contains a description of each method, links to guidance or training and contact details.

Analysis encompasses methods of any type which are used in generating understanding of the processes generating data.  The focus will be on non-routine analyses, and novel methods not yet incorporated into routine ‘pipelines’.  Training needs for more routine analytical methods may be addressed, depending on demand.

There is special interest in Bayesian Methods (where applications may be wider, e.g. including linking, geographic data, appropriate quantification of uncertainty, inclusion of prior knowledge).

 

Guidance

This will be updated as and when guidance is published.

 

Training

Online

For information on courses available please email the Methodology Advisory Service: MAS@ons.gov.uk

Classroom based

Specific training courses can be designed on request.


About this expert group

All the other expert groups almost certainly involve elements of analysis.  These other expert groups mostly represent specific and major areas of existing effort within methodology – linked with methods that are currently used by business areas in gathering or processing data and producing routine outputs.

The analysis expert group does not focus on methods which feature in a major way in other expert groups.

There is also the Data Science and High-performance computing (DaSH) expert group which similarly to the analysis expert group concerns a broad set of techniques, many novel and not (yet) routinely used in ONS. There is a difference of emphasis, with the analysis expert group less concerned with enormous datasets, (fast) data-processing and linking or creating ‘pipelines’, and more concerned with distilling understanding (as opposed to short-term prediction or estimation) than the DaSH expert group.

There will however inevitably be some areas of overlap between expert groups.

The aims of the analysis expert group include:

  • being a forum for identifying, discussing and disseminating of ideas and methods.
  • acting as a clearing house for provision of advice and team formation.
  • identifying and providing training (alongside the GSS Capability Team).
  • providing best practice guidance.
  • being a forum for development of research into methods.
  • being active in making links and inviting participation across GSS and beyond

To get in touch with the Analysis expert group please email Analysis.Expert.Group@ons.gov.uk

Data linking is the process of joining datasets together so that we can make as much use as possible of the information they hold.  The government holds an enormous amount of data. By using it effectively, we provide insight, drive policy change and answer society’s most important questions. While datasets are useful on their own, bringing them together means that we can take advantage of the combined resource they offer.

 

Guidance

Data linking

Communicating quality, uncertainty and change

GUILD: GUidance for Information about Linking Data sets

 

Training

Online

Introduction to data linking

Other related courses can be found on the data linkage page.

Classroom based

For working and practitioner level class-based courses, please contact GSS.Capability@statistics.gov.uk

Specific training courses can be designed on request.

 

About this expert group

The Data Linkage Expert Group provides a forum for sharing knowledge on linkage methods typically used in the production of official statistics and also research on relevant developments. The group also provides advice, develops training and best practice guidance and makes links across the GSS and beyond.

 

To get in touch with the Data linkage expert group please email DataLinkage@ons.gov.uk

Our statistical computing work uses a combination of statistical, survey and programming knowledge to deliver a variety of Information Technology (IT) solutions to support colleagues. Here, we develop and maintain robust statistical production systems which both accurately implement the methods specified by the ONS methodology division and meet user’s requirements, both in terms of functionality and usability.

 

Guidance

We have written a blog post: ‘Happy campers: Using machine learning to identify caravans in Zoopla data

 

Training

Online

For information on courses available please email the Methodology Advisory Service: MAS@ons.gov.uk

Classroom based

Specific training courses can be designed on request.

 

About this expert group

DaSH stands for Data Science and High-performance computing. Founded from a blend of the capabilities in statistical computing and the original ONS Big Data Team we offer expertise in all areas of data science and high-performance programming of statistical methods. We are constantly exploring new techniques and applications and make our work open to the public, where possible.

Complimentary to the ONS Data Science Campus, our focus is more on the application of data science and computing techniques within the production of official statistics and we make use of the wealth of expertise in the ONS methodology division to bring a broad range of statistical theory into our work.

Our projects might use any mix of traditional official surveys and administrative data together with alternative data from the private sector or internet. We ensure that all our work satisfies our obligations under the Code of Practice for Statistics paying specific attention to the reproducibility of our results and maximising transparency of our methods.

To get in touch with the Data science and high-performance computing expert group please email DaSH@ons.gov.uk

Demographic methods are concerned with measuring and analysing the size and structure of the population and how it changes over time. The processes of population change (such as migration) are of particular interest and attention.

Guidance

Enquiries are channelled to relevant experts, so if you have a demographic query you would like guidance on, please contact us. We want to help forge innovation through collaboration across the GSS and beyond.

 

Training

For information on courses available please email the Methodology Advisory Service: MAS@ons.gov.uk

 

About this expert group

The demographic methods expert group provides a forum for giving a cross-cutting steer and overarching view on demographic work, as well as the capacity to look at new and emerging methods. The group has members interested in or experts in demographic methods from across ONS, with plans to extend to the wider GSS. Topic areas include:

  • measuring uncertainty
  • frameworks for managing error
  • blended estimation
  • fertility
  • mortality
  • migration including modelling approaches for international migration

We also provide international peer review and an opportunity for experts to come together outside of project-based constraints. The group involves sharing and discussing any relevant topics, journal articles and providing support in demographic methods.

We are currently developing methodological approaches for statistics in response to the coronavirus (COVID-19) pandemic. We are seeking to develop our understanding of the implications of the current situation for our demographic models and statistics. We are keen to collaborate across the GSS, with academia and internationally on this topic. Please contact us if this is something you can get involved with.

To get in touch with the Demographic methods expert group please email Demographic.Methods@ons.gov.uk

Editing and imputation are both methods of data processing. Editing refers to the detection of errors.  We can correct for errors in the data by applying imputation. Imputation refers to estimating values for missing or inconsistent data items.

Data collected from surveys and alternative data sources always contain various errors which need to be identified and removed or corrected. This improves the data quality, reduces non-response bias and ensures outputs from the data are plausible.

Types of error include:

  • missing values – questions in a survey not being answered
  • invalid values – characters where numbers are expected for example
  • inconsistencies either within or between records, or external data –  for example, a child has been recorded as being older than their parent

 

Guidance

UNECE Generic Statistical Data Editing Model
A reference for all official statisticians whose activities include data editing.

Flexible Imputation of Missing Data
Provides a good overview of imputation including missingness mechanisms and methods, with examples that can be performed in R.

 

Training

Online

Introduction to editing and imputation

Classroom based

We run practitioner-level workshops that are more in-depth and are aimed at those wanting to use specific methods in their own work. These can be tailored to your specific needs. Previous class-based training has been delivered on methods such as donor and model based imputation, and automatic and selective editing.

 

About this expert group

The editing and imputation expert group provides a forum for sharing knowledge on editing and imputation methods typically used in the production of official statistics and also research on relevant developments in these methods. The group also provides advice, develops training and best practice guidance and makes links across the GSS and beyond.

 

To get in touch with the editing and imputation expert group please email Editing.and.Imputation.expert.group@ons.gov.uk

Index numbers are used to measure changes and simplify comparisons. ONS produces index numbers principally in the field of economics. Economists are interested in how changes in the monetary value of economic transactions can be attributed to changes in price (to measure inflation) and changes in quantity (to measure sales volume or economic output). Index numbers typically measure these changes over time.

 

Guidance

Government Statistician Group List A Topics (PDF, 353KB)

Index numbers page on the ONS website

 

Training

Online

For information on courses available please email the Methodology Advisory Service: MAS@ons.gov.uk


About this expert group

The index numbers expert group is a voluntary group of methodologists and analysts with mixed experience, who share an interest in index number theory. The expert group is involved in a variety of new developments related to index numbers, providing guidance on best practice and appropriate methods.

The expert group aims to:

  • provide a source of expertise on the formulation of index numbers and related topics
  • provide advice and quality assurance to producers of indices
  • carry out research into new methods
  • provide training on index numbers to ONS and GSS staff

To get in touch with the index numbers expert group please email Index.Numbers.Expert.Group@ons.gov.uk

Qualitative methods are often used in primary research with the public to understand societal constructs, attitudes, and behaviours. Methods of data collection include cognitive and depth interviews, focus groups, observational studies, and open survey questions.  They result in textual qualitative data which are analysed using thematic analysis, or more recently cognitive mapping and natural language processing.

Data collection methods are (mainly) qualitative methods used to help design robust survey questionnaires and help to understand and address sources of measurement and other non-sampling error, exploring cognition, respondent burden, usability, accessibility, mode effects and response errors. They can be used in combination with quantitative data such as coded survey, observational (e.g. usability) or paradata to understand how respondents interact with surveys.

Guidance

We provide ad hoc guidance upon request and plan to publish some guidance in the future, for example around remote qualitative testing.


Training

Online

For information on courses available please email the Methodology Advisory Service: MAS@ons.gov.uk

Classroom based

MSc in Data Analytics for Government

For working and practitioner level class-based courses, please contact GSS.Capability@statistics.gov.uk

Specific training courses can be designed on request.


About this expert group

Members of qualitative and data collection methods (QDCM) are mainly social research methodologists. QDCM provides expertise on qualitative and data collection methods and analysis. We work with customers in survey transformation, survey production and statistical analysis/output areas, within and external to the Office for National Statistics, to help design, evaluate, and implement robust social and business survey questionnaires including the Census. This includes interviewer-administered (face-to-face and telephone) and self-administered questionnaires (online, paper), mixed-mode surveys, other collection tools and respondent materials such as record-keeping diaries, letters and leaflets.

Our work includes the following elements, as appropriate to each customer’s needs:

  • collaboration with partners and customers to understand and establish clear data requirements and research questions
  • exploring survey concepts and design aspects (which might include depth interviews/focus groups)
  • developing draft survey questions using good practice principles, and based on experience
  • conducting an expert review of survey questions, to identify potential sources of error and propose amendments
  • conducting cognitive question testing, to explore whether questions will collect data consistent with data user requirements, and that respondents comprehend and are able and willing to answer questions
  • conducting user testing to ensure respondents are able to use collection tools without difficulty
  • carrying out wider qualitative research projects, such as consultations and feasibility testing
  • advising on and quality assuring data collection and qualitative research design including purposive sampling and research ethics
  • upskilling our clients in qualitative and data collection methods
  • providing training on qualitative and data collection methods
  • producing research on new/developing qualitative and data collection methods, for example online research methods, using Natural Language Processing in analysis
  • carrying out research with vulnerable people

To get in touch with the qualitative and data collection methods expert group please email DCM@ons.gov.uk

Sample design covers all aspects of how the samples in business surveys are specified and selected.

 

The design of samples is important as it provides the basis for sound measurement of economic  phenomena from surveys of businesses in the UK.

 

Estimation is also known as weighting. It describes the method by which we produce estimates about the characteristics of the population as a whole from the responses given by those businesses selected in the sample survey. Responses from the sample are weighted to ensure they represent the entire population without bias and produce good quality outputs.

 

Guidance

This section is awaiting content.

 

Training

For information on courses available please email the Methodology Advisory Service: MAS@ons.gov.uk

 

About this expert group

Business surveys select a sample of businesses, rather than surveying them all like a census. Our expert group ensures the sample is representative of all businesses, and that sufficient data are gathered from these businesses and the right methods are used to accurately and precisely estimate statistics for all businesses.

To get in touch with the Sample design and estimation for business surveys expert group please email Sample.Design.Estimation.Business@ons.gov.uk

Sample design covers all aspects of how the samples in social and population surveys are specified and selected. The design of samples is important as it provides the basis for sound measurement of social phenomena from surveys of households in the UK.

Estimation for social surveys is also known as weighting. It describes the method by which we produce estimates about the characteristics of the population as a whole from the responses given by those households selected in the sample survey. Responses from the sample are weighted to ensure they represent the entire population without bias and produce good quality outputs.

 

Guidance

This section is awaiting content.

 

Training

For information on courses available please email the Methodology Advisory Service: MAS@ons.gov.uk

 

About this expert group

The sample design and estimation for social surveys expert group provides a forum for sharing knowledge on sampling and estimation methods typically used in the production of official social statistics and also research on relevant developments. The group also covers methods for estimating population size as part of the decennial census. The group also provides advice, develops training and best practice guidance and makes links across the GSS and beyond. To get in touch with the Sample design and estimation for social surveys expert group please email Sample.Design.Estimation.Social@ons.gov.uk

Most government surveys are designed to provide statistically reliable estimates at national or high level geographies. So, when statistics are required for more detailed geographical breakdowns (small areas) or small subgroups (domains) of the population the sample sizes just aren’t large enough to produce robust estimates.

Increasing the size of samples would be prohibitively expensive – instead small area estimation methods have been developed that combine data from administrative, census and survey sources to produce estimates for small areas or domains.

There are many statistical techniques covered by small area estimation – a frequently used approach is a model-based one, where local area outcomes are estimated using a multilevel regression of survey data on auxiliary data such as census and/or administrative data sources.

At ONS there is currently a drive to produce estimates without the use of the census. Small area estimation provides a useful framework for this since it naturally allows survey data and administrative data to be combined in estimation of the precise quantity of interest.

 

Guidance

ESSnet Guidelines for the application of the small area estimation methods in NSI sample surveys (PDF, 935KB)

ESSnet on small area estimation

 

Training

Online

For information on courses available please email the Methodology Advisory Service: MAS@ons.gov.uk

 

About this expert group

This expert group provides a forum for sharing knowledge on these methods typically used in the production of official statistics and also research on relevant developments. The group also provides advice, develops training and best practice guidance and makes links across the GSS and beyond.

 

To get in touch with the Small area estimation and modelling expert group please email Small.Area.Estimation.Modelling@ons.gov.uk

Statistical Disclosure Control (SDC) is the application of methods to reduce the risk of disclosive information about data subjects. Methods for SDC usually restrict the amount of or reduce the detail of the data released.

SDC modifies data so that the risk of data subjects being identifiable is within acceptable limits. The aim is to make the data as useful as possible, while maintaining appropriate confidentiality.

An alternative approach is to enable access to unaltered data to a small number of users under strict licensing conditions. Outputs from all analyses will be checked to ensure they are not disclosive.

 

Guidance

Disclosure control for tables produced from surveys

Disclosure control for tables produced from administrative sources

Disclosure control for microdata produced from social surveys

Privacy and Data Confidentiality National Statistician’s Quality Review December 2018

 

Training

Online

For information on courses available please email the Methodology Advisory Service: MAS@ons.gov.uk

 

About this expert group

This expert group provides a forum for sharing knowledge on disclosure control methods typically used in the production of official statistics and also research on relevant developments. The group also provides advice, develops training and best practice guidance and makes links across the GSS and beyond.

 

To get in touch with the Statistical disclosure control expert group please email SDC.Queries@ons.gov.uk

The data we work with often comes in the form of a time series, i.e. a series of data points representing values at different points in time. This section provides links to the time series analysis help available to the GSS.

 

Guidance

ESS Guidelines on Seasonal Adjustment

ESS Guidelines on Temporal Disaggregation

Handbook on Seasonal adjustment

Eurostat Time Series Page on CROS

A brief description of JDemetra+

Exploring the effect of weather and climate on official statistics: Guidance for official statistics producers

ONS guide to seasonal adjustment with X-13ARIMA-SEATS

 

Training

Online

Introduction to time series

Introduction to seasonal adjustment

Introduction to forecasting

One-hour introduction to seasonal adjustment (this is a live Skype session) –  This session repeats each month, with a trainer going through introductory slides live on Skype. This is still a pilot and currently only available internally to the Economic Statistics Group (ESG) but will be made available more widely in the future.

 

About this expert group

This expert group provides a forum for sharing knowledge on time series methods typically used in the production of official statistics and also research on relevant developments in time series methods.

Time series methods commonly found in the production of official statistics include:

  • seasonal adjustment – identifying, estimating and removing regular seasonal and other calendar related movements from time series
  • temporal disaggregation – estimating higher frequency time series from lower frequency time series possibly using indicator series
  • benchmarking – adjusting higher frequency time series to be temporally consisted with a lower frequency time series
  • forecasting – estimating future values of a time series based on past values
  • revisions analysis – testing revisions for evidence of bias as updated versions of data become available
  • interpolation – the process of estimating and inserting missing values in time series data
  • other time series modelling – for example, models to interpret movements in time series (i.e. AutoRegressive Integrated Moving Average (ARIMA) models) or estimate discontinuities (i.e. state space models)

To get in touch with the Time series expert group please email TSAB@ons.gov.uk

 

Contact

For further information or advice please email the Methodology Advisory Service: MAS@ons.gov.uk.