TEDS Data Dictionary

Sample Sizes for TEDS Studies

This table summarises the sample size, in terms of number of families initially contacted, for each of the main TEDS studies. It does not include TEDS studies that used smaller subsets of the cohorts (e.g. 4 Year In Home, post-18 spatial web studies, post-21 g-game and covid studies, post-26 CATSLife study). Nor does it include secondary data collections from the main studies (e.g. LEAP-2 at age 16, TEDS21 phase 2). The figures are based on records kept in the TEDS admin system.

The sample sizes shown in the table only include families in the TEDS sample, not in the original ONS sample. As mentioned above, between 1st Contact and the 9 Year study, there were continuing attempts to contact those families in the wider ONS sample that had not yet responded, and were therefore not yet enrolled in TEDS. These 'inactive' families were removed from the TEDS sample after the 9 Year study. Inclusion of only subsets of the TEDS sample in this table provides comparability across all TEDS studies.

Some attempts at initial contact inevitably fail (generally because some participants change address, phone or email details without notifying TEDS), so the quoted figures are likely to be slight overestimates of the actual number of contacts made.

In the later studies, starting with the 21 Year and increasingly thereafter, there were families in which only one or two of the three family members (parent + two twins) could be contacted. In some cases, the parent or one or both twins independently has withdrawn from TEDS. In other cases, twins moved away from the parental home then either parent or twin became uncontactable. The figures in the table below include all families in which at least one family member was contacted. A column has been included to show how many of the contacted families included only one contactable twin.

For more information about the families not contacted, see the following section about families excluded from TEDS studies.

Families Excluded from TEDS Studies

This table summarises the numbers of families omitted from TEDS study samples for various reasons. As above, for the sake of simplicity, the table includes all major TEDS studies but not those studies that included smaller subsets of families, nor secondary data collections from the main studies.

Based on the surviving records, especially for the earlier studies, it is not always possible to recreate the exact reasons why particular families were not contacted. For example, the dates when some families withdrew are unclear, as are the dates when some families became address problems. Furthermore, the admin records are likely to contain occasional errors in the logging of items sent to and received from families. The numbers in this table are therefore approximate, and have been rounded to the nearest 50 families.

As above, for better comparability across all TEDS studies, the numbers only include families in the TEDS sample. In other words, with the exception of the 1st Contact study, the numbers do not include those inactive families in the ONS sample who were later removed from the TEDS sample, even if there were attempts to contact them between 1st Contact and the 9 Year study.

The table above illustrates the following trends, contributing to overall declines in sample sizes over time (see the glossary for further explanations of terms):

• Withdrawn families. The number of withdrawn families increases steadily over time. With each study, more families tell us that they wish to withdraw, for a variety of reasons including lack of time, and twins losing interest as they grow older. Withdrawal is permanent and irreversible for participants.
• Address/contact problems. The number of address problems has also increased steadily over time. More address problems emerge with each postal mailing, and only a minority can be traced. The apparent decrease in address problems from 18 year to 21 year was caused by the deliberate withdrawal of a number of long-term uncontactable and unresponsive families (hence also an increase in withdrawn families). The decrease from 21 year to 26 year may also partly be explained by withdrawals, but may also be a result of the increasing use of twins' own contact details (rather than those of their parents), with more reliance on email rather than postal contact.
• Medical exclusions. The number of "medical exclusion" families has remained fairly small (a few hundred) but has varied over time for several reasons. Firstly, during twin childhood and adolescence, there was a general increase as families informed TEDS of severe medical conditions that were previously not recorded for the twins. Secondly, the definition of what types of medical condition should be excluded has been modified from one study to the next. The redefinition of medical exclusions has varied to some extent according to the demands of a particular study, and judgments by TEDS staff as to which twins might find activities difficult (or in some cases which twin pairs might be excluded from analysis even if they provided data). During the 18 year study, a final attempt was made to involve many medical exclusion families who had been very unresponsive in recent studies; those who did not respond were withdrawn, hence from the 21 year study onwards there was less reason to remove medical exclusions from samples. Then, after the completion of the TEDS21 study, in which twins themselves were asked about their medical conditions, the entire categorisation of medical exclusions was overhauled and was defined per-twin instead of per-twin-pair; this resulted in a significant reduction in the number of twins treated as medical exclusions. See the exclusion page for further information.
• No recent data. A family that chooses not to participate in one TEDS study may or may not choose to participate in the following study. This pattern is unpredictable but, over time, a family that has chosen not to respond in successive studies is less likely to respond when asked again. The table shows that there have been attempts to eliminate such unresponsive families from some study samples (notably the 4, 10 and 12 Year studies), with the aim of increasing participation rates.
• Email addresses. From age 18 onwards, twins have been increasingly contacted by email (either in addition to, or instead of, by post). Until age 18, initial invitations were generally made by post, but since age 21 initial invitations have been sent by email before following up with postal invitations. The existence of twin email addresses is linked to past twin activity, because email addresses have been collected as part of twin studies from age 16 onwards.

Attrition in TEDS

In the context of this examination of TEDS study samples, attrition means the reduction, over time, of the number of contactable families (and hence twins) within the TEDS sample. The main causes of attrition in TEDS are withdrawals from the study, and loss of correct contact details (address problems and email problems).

Another cause of attrition is that, over time, increasing numbers of participants simply decline to respond, without explicitly withdrawing from TEDS. This type of attrition is difficult to measure and is not examined further on this page, which has a focus on the numbers of contactable and contacted participants. The data returns page provides information about changes over time in the numbers of responding participants.

Withdrawal is irreversible: it means a family, or an individual participant, chooses to leave the study and not be contacted again. As shown above, the number of withdrawals inevitably increases over time, leading to a decrease in the contactable sample size.

Loss of correct contact details, usually through address problems, may be reversed for any individual family or participant through tracing. However, tracing proves to be much harder for some participants than others; it becomes harder when a participant has been out of touch for a long time. As a result of these difficulties, some participants remain apparently permanently untraced, leading to an increase over time of the number of uncontactable families.

Based on the surviving records for earlier TEDS studies, we cannot directly measure the numbers of contactable families retrospectively. However, we do have good records of the numbers of families actually contacted. In studies where there were no other factors affecting selection, we can build a picture of how the number of contactable families has decreased over time. The table below shows the numbers of apparently contactable families in the entire TEDS sample at 1st Contact compared with the 7, 8, 14, 16, 18, 21 and 26 Year studies.

Additionally, the 2, 3 and 9 Year studies can be compared with the rest by looking at the number of families contacted within school cohorts 1 and 2, which were subsets of all the other study samples. School cohorts 1 and 2 are families with twins born between January 1994 and August 1995.

The table excludes the 4, 10 and 12 Year studies, in which families were selected for contact on the basis of previous activity, and in which the actual number of contactable families is now very difficult to measure.

Where possible the figures have been adjusted so as to include families categorised as medical exclusions, if apparently contactable, even if they were not actually contacted in a given study.

The table shows that the number of contactable families has fallen steadily over time. The only exception seems to have been a modest increase between the 21 Year study and the 26 Year study, over the entire sample (although not for cohorts 1 and 2). As mentioned above, this seems to have been because many more twins were contacted independently at age 26, using their own email and/or postal addresses, and this included some families that had been uncontactable at the parent postal address at earlier ages.