TEDS Data Dictionary

Parent questionnaire returns

This table includes all the main TEDS studies in which parent data were collected in bulk by means of a booklet or questionnaire (often sent together with twin booklets). Note that at age 21, parents could use paper booklets but they could also use the web or a phone app to return the same data; at earlier ages, paper booklets were generally used.

This table clearly shows an overall decline in the % return rate, but with variations possibly due to the following factors:

• Removal of inactive families. After the 9 year study, "inactive" families were removed from the TEDS sample (see inactive families). Removal of these unresponsive families from the contacted sample inevitably boosted the % return rate from age 10 onwards.
• Sample selection. Samples selected on the basis of recent data returns are likely to give a higher % return rate. This will have affected the returns at ages 3, 4 and 12.
• Telephone contact. Higher response rates may be observed when families are contacted by telephone, than when the contact is entirely via mail. There were extensive telephone contacts with families during the 7, 12 and 21 Year studies.
• Prior consent. In the 9 Year study, written consent was sought before booklets were sent (in other studies, consent forms were generally sent and returned simultaneously with the booklets). This two-stage process of data collection may have resulted in a drop in overall return rate.
• Other factors that may have affected returns, but whose effects are difficult to judge from the raw returns figures, are the nature and length of the questionnaire; the timing of sending the questionnaire; the number and frequency of reminders sent; other simultaneous demands placed on families (e.g. web or phone tests); the method of booklet data collection (paper and/or web and/or app); and whether or not vouchers or other rewards were simultaneously offered for the return of twin data.

Twin questionnaire returns

This table includes all the main TEDS studies in which twin questionnaire data were collected in bulk (as distinct from twin cognitive test data below).

For the purposes of this table, the number of twin pairs contacted includes some twin pairs in which one twin was contactable (unless otherwise stated, such pairs were a small minority). For the later web questionnaires, the number of twin pairs in which both twins finished only includes those pairs in which both twins finished the entire questionnaire (others are included in the figures for partial data).

At ages 9, 12 and 16 these were exclusively paper questionnaires. At age 14, paper booklets were used but with the option of using the web. The 18 Year Fashion questionnaire involved cohort 3 only, and was delivered exclusively via the web. In TEDS 21, twins had the options of using a phone app, the web or a paper booklet to return the same questionnaire. In the 21 year covid study, the only option was a web questionnaire. In TEDS26, most twins returned data via the web but there was the option of a paper version.

* The exam results questionnaires at ages 16 (GCSEs) and 18 (A-levels) were collected from parents as well as from twins. Where twins could not be persuaded to return the questionnaires on paper, parents were called and were asked to give the twins' exam results by phone. Hence a significant portion of the data returned was from parents.

** In the covid study, in phase 1, 1271 unpaired twins were contacted in addition to the 6673 pairs. However, most data returns came from the latter group. In phases 2, 3 and 4, the contacted sample was reduced, removing twin pairs who had not responded recently, leaving fewer than 300 unpaired twins who were contacted.

The % return of complete data has fluctuated greatly, due to a variety of factors discussed above and below. However, the pattern in the % return of partial data is clearer: this has increased significantly, especially since the introduction of electronic data collection (web or app).

Some of the factors affecting parent questionnaire data returns (see above) will also apply here to the twin questionnaire data returns. In fact, in the booklet studies at ages 9, 12, 14 and 16 the twin pair returns are almost identical to the parent data returns; this was because the parent and twin booklets were generally sent and returned as a set together. The following additional factors may also have contributed to variations in the % data return rate for the twin questionnaires at later ages:

• Electronic data collection. The use of the web (and later a phone app) to collect questionnaire data relies increasingly on independent participation of each twin, hence the increasing % of partial data. This first became evident at age 14 (where some but not all twins used the web) then much more so at ages 18, 21 and 26.
• The special case of exam results. The exam results data collections (GCSEs at 16, A-levels at 18) were unusual in that the content was factual and brief, could be collected from parents if twins did not respond, and could be collected entirely by phone in a short interview. These factors largely account for increased data returns.
• Nature and length of questionnaire. The exam results questionnaires, the Fashion questionnaire and the covid questionnaire were all short, contributing to improved data returns. The nature of these questionnaires may also have been more appealing to families, who generally seemed very willing to share exam results, and who may have been keener to give their opinions on fashion/food/music and the covid crisis than to answer the more conventional behaviour questionnaires.
• Time scale for data collection.
  The Fashion study data collection, and each phase of covid questionnaire data collection, was restricted to roughly a month, giving little time for the usual reminders. Furthermore, in the covid study only email contacts were used, as the circumstances did not allow for sending postal invitations or for telephone calling. The Fashion and covid questionnaires were similarly brief (roughly 15 minutes to complete). These factors make them difficult to compare with other studies. However, possible reasons why data returns were higher in Fashion than in the covid study could include voucher rewards (given for the former but not the latter) and age effects (the covid study took place roughly 5 years after the Fashion study).

Twin test data returns

This table includes the main TEDS studies in which twins were asked to complete cognitive and other tests, either by telephone (at 7) or via the web (at 10, 12, 14, 16 and 18). Parents were asked to give consent before twins started the tests, up to age 16; both written and verbal consents were used at 7, while on line consent was required before the web tests could be started at 10, 12, 14 and 16 (written and verbal consents were also sought at these ages, but these were additional to the on line consent). From age 18 onwards, no parental consent was sought because twins were adults; on-line twin consent was built into the start of the tests themselves, and is not shown separately in the table below.

* In each of these studies, over 1000 unpaired twins were contacted in addition to the invited pairs; but a large majority of the data returns come from the invited pairs.

Some of the factors affecting parent data returns (see above) will also apply here to the % parental consent rate. For example, sample selection (particularly at 12 and 18, and to a lesser extent at 10); and the degree of telephone contact with families (used extensively at 7, 10, 12 and 16, but much less so at 14 and not at all at 18).

Similarly, some factors affecting twin questionnaire data returns (see above) will also apply here to twin test data returns. For example, longer batteries of tests may be linked to lower data returns (the web batteries at 12 and 16 were much longer than those at 10 and 14, while the Kings Challenge and Navigation batteries were longer than those of Perception and Bricks). The 21 Year g-game was short, but on the other hand the time scale for data collection was short, with few reminders, and there were no twin rewards on offer apart from a prize draw.

The following additional factors may also have contributed to variations in the % data return rate for the twin cognitive studies:

• Voucher rewards. These were introduced in the 10 year study, and were used again at 12, 16 and all the 18 year studies, to reward twins for completing the web tests. However, vouchers were not offered in the 14 year study, except in cohort 1. This helps to explain the significant drop in returns of twin data at age 14 compared with most studies before and after. Vouchers were not offered in the 21 year g-game (except in a prize draw)
• Twin age. The difference between the consent rate and the data return rate was narrower at 7 than at the later ages; this perhaps reflects increasing levels of twin independence with age. Increasing twin independence may also help to explain the generally increasing % of partial data returns.
• Internet connections. At the time of the 10 year study, a significant number of families had no internet connection at home, or had slow dial-up internet connections. By the time of the 12 year and later studies, home computers with broadband connections were becoming the norm. This may have been a factor contributing to the relatively low % data return rate at age 10.
• Technical problems and twin fatigue. Many twins experienced technical problems with the Navigation activities, and to a lesser extent with the Kings Challenge activities, and this may have discouraged participation and led to an increase in partial data returns. Furthermore, broadly the same sample of twins (subsets of cohorts 1 and 2) were invited to take part in Perception, Bricks, Kings Challenge and Navigation, the latter three studies all starting within the same year; this may help to account for the drop in returns in the latter two studies, even though the activies were "gamified" in an attempt to make them more visually appealing.
• Platform restrictions. Some online activities could only be completed using particular types of devices, or with specific browser versions or other software. This affected returns in the early web studies, when some families did not have home computers; it affected the Navigation study, which required updated browsers and specific plugin software; and it affected the CATSLife study, in which some of the activities could only be completed on a laptop or desktop, not on mobile phones.

Teacher questionnaire returns

Collection of teacher data at each age involved several steps: firstly, the parent was asked for consent for TEDS to contact the twins' teachers; at the same time, the teachers' names and school addresses were sought; then questionnaires were sent to teachers. In some cases, twin pairs were taught in the same classroom, so the two questionnaires could be sent simultaneously to the same teacher; usually, the two questionnaires were then returned in the same envelope. In other cases, the two twins were separated at school, so the two teacher questionnaires had to be sent and returned independently.

In primary schools, the contacted teacher was usually the teacher who taught the twin's class full time, for most or all subject areas. However, in secondary schools, where children generally received specialist subject teaching from a range of teachers, the contacted teacher was usually the form teacher (who took registration and had daily contact with the children).

The % data return rate was significantly lower from secondary school teachers (approx. 65-70%) than from primary school teachers (approx. 75-85%). Feedback from some form teachers that were contacted in secondary schools indicated that they did not always know the twins well enough to be able to fill in the questionnaires confidently; they were often not the twins' subject teachers, and were frequently unable to fill in the academic achievement part of the questionnaire. Other variations in the % return rate may be related to factors such as the timing and frequency of reminders sent to teachers, the length of the questionnaire (much longer at 9 than at 7), and so on.

At primary school, the majority of twin pairs had the same teacher for both twins; the teacher questionnaires were therefore usually returned together, and there were relatively few twin pairs (approx. 5-10%) for whom only one teacher returned data. However at secondary school, the majority of twin pairs were split up, and the two teacher questionnaires were usually returned independently by two separate teachers for each twin pair. Hence the increased proportion (approx. 25%) of twin pairs with teacher data returned for only one twin.

Variations in the % parental consent rate are similar to the variations in % return rate for the parent data, and to the variations in % consent rate for the twin tests, as described above. Sample selection was probably an important factor in the increased consent rate at 12, and to some extent at 10. The amount of telephone contact also varied considerably: at 10 and 12, and to a lesser extent at 14, many parents were telephoned to collect teacher consent details, whereas at 7 and 9 these were collected in writing by mail.