Multi-Wave Survey

If a respondent shall be surveyed at different points in time, you are dealing with a multi-wave survey, a longitudinal study, or a diary study. This raises two challenges:

1. How do you get the respondent to fill out the right questionnaire at the right point in time?
2. How are the different records of a respondent matched with each other at the end?

SoSci Survey offers a number of functions that enable or facilitate the implementation of multi-wave surveys.

Technically, there are two ways to implement a multi-wave survey with SoSci Survey. Both have advantages and disadvantages. For both variants, you should test them thoroughly in advance – for example, with shortened intervals (5 minutes instead of 7 days) and disposable email addresses.

The first option is that only one questionnaire is created under Compose Questionnaire. It first contains the pages for the first survey wave, then an interruption page (Pausing an Interview), and then the pages for the second survey wave.

At first, the interruption page shows the respondent a farewell (for the first part of the questionnaire). If the respondent then returns to the questionnaire after a defined period of time (which is implemented, for example, by means of mailResume() or a question of the type Send Email to a Personal Contact), the interruption page directly forwards to the next page, where the second questionnaire starts. Programming an interruption page is documented in the reference for the command mailResume().

• Advantages
  • All of the respondent's data is stored in one record/case/interview.
  • In the second part, the answers from the first part can easily be accessed by means of value() or replace() – e.g., for filter questions or for adjusting questions to the respondent's answers by means of placeholders.
• Disadvantages
  • If questions occur in both parts of the questionnaire, they also have to be duplicated in the list of questions (otherwise, the earlier answers to the question are shown as preselection and then overwritten).
  • It is a little more complicated to specifically send reminder mails to those who have received an invitation to the second part of the questionnaire but have not participated yet.
  • If someone drops out while filling out the first part of the questionnaire, you have to make sure that this person is not invited to the second part – otherwise (without further filters), the person continues filling out at the point where they dropped out.
  • The progress bar has to be adjusted via option('progress', ...) or disabled completely, because otherwise it counts across both parts of the questionnaire.
  • In surveys with more than two survey waves, it becomes complicated as soon as a respondent wants to skip one of the survey waves and participate again later. For this case, additional filters are required in the questionnaire.
  • Once the questionnaire is in the field, no pages may be added to or deleted from the first part of the questionnaire (otherwise the respondent ends up on the wrong page when continuing with the second part).

In diary studies, usually the same questions are asked in each survey wave – and the data is often to be analyzed by means of multilevel analyses. A solution with multiple questionnaires is preferable if…

1. the same questions shall be used multiple times
2. more than two survey waves are planned
3. the data shall be stored in separate records per interview (and not per respondent)

If you use multiple questionnaires, a separate record (data row) is stored for each completed questionnaire. So that these can be matched with each other in the analysis, all data rows require a unique participant code. You can obtain this participant code, for example, as follows:

1. If you send mailings, SoSci Survey automatically marks all records of an address entry with a unique person ID in the variable SERIAL (Additional Variables in the Data Set). This is also possible without losing anonymity.
   • Important: The placeholder %link% must be used in the mailing, so that personalized links are actually sent.
   • Important: All questionnaires must be created within the same survey project; the participant codes are project-specific.
2. You can send personalized links yourself. The respondent's ID is then passed as a reference or as an access code (URL to the Questionnaire). A reference is stored in the variable REF, an access code (which also allows resuming the questionnaire) in the variable SERIAL.
3. The respondents shall create a personal code from various pieces of information (first letter of the mother's first name, last digit of the year of birth, …) and enter it into a text input field. Because of its susceptibility to errors, we generally advise against this solution – and if you use it, it is smarter to ask for the individual components in separate text fields.

Compared to a single questionnaire that is filled out in stages (above), the variant with separate questionnaires has the following advantages and disadvantages.

• Advantages
  • It is (technically) unproblematic if a respondent misses/skips individual survey waves.
  • Sending reminder mails is quite easy.
  • A question can be used in different questionnaires.
  • The same questionnaire can be sent to a respondent any number of times (e.g., diary study)
• Disadvantages
  • If you want to refer to the respondent's earlier answers in later questionnaires, you have to use the Content Database, see Using an answer in later questions.
  • You get several records (data rows) per respondent, which may have to be merged in the analysis.

Depending on the study design, all respondents receive the invitation to the second survey wave at the same point in time – or the point in time depends on when they filled out the first questionnaire (a random assignment of the point in time is also possible, see MESM below).

Sending an invitation at a predefined point in time is done directly via the user interface:

• Under Invitation Mailings → Mailings, a new mailing is created.
  • If respondents were already invited to the first part of the questionnaire via a mailing, select this mailing in the tab Reminder/follow up for Reminder/follow up for (Type of follow-up: “New survey wave”) – then you can send the invitation specifically to those who also filled out the first questionnaire.
  • In the tab Questionnaire URL, select the correct questionnaire.
• With the button Sending the Mailing, the dispatch is then prepared.
  • As Time of dispatch, simply enter the point in time when the invitation shall be sent.
  • Under Recipient selection: reminder/follow-up mail, you can additionally specify, if necessary, that only addressees who have filled out the first part of the questionnaire receive an email (see above).

As a variant of the uniform dispatch time for all address entries, individual points in time can also be specified manually per address entry: Import Sending Times

If the invitation to the second part of the questionnaire shall be sent depending on when the first part was filled out, a little PHP code comes into play (see also Introduction to PHP). The function mailSchedule() specifies when a mailing shall be sent to a certain respondent.

Ideally, the respondent is already invited to the first questionnaire by means of a mailing. The procedure if the email address is not known in advance is described in the chapter Multi-Wave Surveys with Self-Recruited Respondents.

In the first questionnaire (preferably towards the end, on a page that dropouts do not reach), the command mailSchedule() is then placed as PHP code, e.g.

mailSchedule(false, 2, 7 * 24 * 3600);

The first parameter false means that a mailing shall be sent to the respondent who is currently filling out the questionnaire. The second parameter 2 specifies that the mailing with the number 2 shall be sent. The third parameter specifies that the mailing shall be sent in exactly 7 days, given in seconds (an hour has 3600 seconds, so 7 * 24 * 3600 gives the number of seconds for 7 days).

Important: Each mailing can be sent to each respondent only once. If you would like to send the same email several times, create a corresponding number of copies of the mailing. The corresponding PHP code could then look as follows (one mailing every day for 5 days):

mailSchedule(false, 2, strtotime('+1 day'));
mailSchedule(false, 3, strtotime('+2 day'));
mailSchedule(false, 4, strtotime('+3 day'));
mailSchedule(false, 5, strtotime('+4 day'));
mailSchedule(false, 6, strtotime('+5 day'));

The PHP code above uses the function strtotime(). It creates a Unix timestamp (number of seconds since 1.1.1970) from a text specification. So '+1 day' becomes the timestamp for “tomorrow at the same time”. '+1 day 8:00' would become the timestamp for “tomorrow at 8 o'clock in the morning”.

If you work with a fixed dispatch date, a reminder mail can easily be created and sent as a new mailing.

If, on the other hand, you work with relative times, you have to prepare both the invitation to the second questionnaire and the reminder mail in the first questionnaire by means of mailSchedule(). This means that, without further action, both mailings would be sent.

mailSchedule(false, 2, strtotime('+7 day'));  // Invitation to the second questionnaire
mailSchedule(false, 3, strtotime('+8 day'));  // Reminder mail one day later

In the second step, you have to make sure that the reminder mail is not sent if the respondent fills out the second questionnaire. To do so, place the command mailRevoke() in the second questionnaire. This function removes the scheduled dispatch from the schedule again.

mailRevoke(false, 3);

Multi-wave surveys allow very varied survey designs. This section is dedicated to some special use cases.

For multi-wave surveys, it is an advantage if you know the respondents' email addresses. This allows you to send them a (personalized) invitation to the second, third, … part of the questionnaire. If the email addresses are not known in advance, the following chapter describes a possible solution: Multi-Wave Surveys with Self-Recruited Respondents

An alternative is to ask for a personal code in each questionnaire, by means of which the records from the different survey waves can later be matched with each other: Request of a Personal Code

It is also possible to actively give the persons a code in the first survey (Display Individual Codes or Voucher Codes), which they then have to enter again in each survey wave. Or, if you already have contact with the persons in advance, you can also hand them such an access code by letter or as a slip of paper at an in-person event. If you use Access Codes for this, SoSci Survey automatically checks the validity of the code.

In a diary study, you usually create three questionnaires:

• Pre questionnaire
• Questionnaire to be filled out daily
• Post questionnaire

If you do not know the respondents' email addresses in advance, you may need a fourth questionnaire for registering the respondents (Multi-Wave Surveys with Self-Recruited Respondents).

A mailing is then created for each email and reminder mail that each respondent shall receive. E.g., two mailings for the pre questionnaire (invitation + reminder each) and one per day. In the list of mailings, there is a symbol for quickly creating copies of a mailing ().

Then, for each mailing individually, you specify by means of Sending the Mailing when it shall be sent.

Tip: Which address entries the mailing is sent to is only determined at the time of dispatch – if it is a uniform point in time for all addresses. So addresses can still be added after the point in time has been specified.

Important: Use the placeholder %link% or %link.html% in the mailing, and select the questionnaire to which the mailing shall refer in the tab Questionnaire URL. The placeholder ensures that the data rows each carry the person ID in the variable SERIAL.

A special form of the diary study is mobile experience sampling. Here, text messages (SMS) are sent to the respondent's smartphone at random points in time (within predefined time windows), asking the respondent to fill out a short (!) questionnaire as soon as possible.

• SoSci Survey also supports sending SMS in the mailing function.
  • In mailings (if the SMS function is enabled), you can leave the text and HTML version of the mailing empty (optional) and enter the text for an SMS instead. If texts for email and SMS are entered, SoSci Survey sends the message via all channels available for the respective address entry (i.e., email and/or SMS).
  • In the list of contacts, mobile phone numbers can be stored instead of (or in addition to) email addresses.
  • The opt-in question can ask for a mobile phone number in addition to an email address and send an SMS for the double opt-in.
    Note: If an email and an SMS are sent for the opt-in, the address entry only has to be confirmed on one of the two channels. In this case, it cannot be verified whether the second contact option is correct.
  • Important: Mobile network operators charge a fee for sending SMS; costs of approx. 0.10 € (plus VAT) are incurred per SMS.
  • Important: For sending SMS outside the EU (plus Switzerland), additional registrations or additional information in the questionnaire are partly required. There may also be additional costs, for example monthly fixed costs for sending to the USA. Please allow 4 weeks of lead time if you would like to send SMS outside the EU.
• SoSci Survey supports sending mailings (or SMS) at random points in time.
  • When sending a mailing (Send Mailings), you can select the option “time-randomized dispatch” for Time of dispatch and then define the period.
  • Important: A separate mailing has to be created for each invitation the respondent shall receive. For example, if 3 SMS per day shall be sent over 7 days, 21 mailings would have to be created (mailings can easily be duplicated). For the first mailing, a random dispatch on Monday between 8 and 11 a.m. would then be scheduled, for example, for the second mailing a random dispatch on Monday between 11 a.m. and 2 p.m., and so on.
  • Important: The time-randomized dispatch assigns a specific point in time to each respondent. If addresses are added to the list of contacts later, they do not receive the mailing, because no dispatch time has been specified for them.
• Questionnaires created with SoSci Survey can usually also be filled out on a smartphone – but for an MESM study, it is worth optimizing the questionnaire specifically for smartphones (Questionnaires for Smartphones), among other things with the corresponding questionnaire layout

The counterpart to experience sampling is that respondents open a questionnaire on their own when a certain event occurs. Ideally, they then do not have to enter a personal code each time, but can directly use a personalized link. There are several ways to do this:

• You can send personalized links by means of a mailing (Send Mailings) that allow filling out a questionnaire multiple times. To do so, set the option “Multiple completion with the personalized link allowed” for the setting Multiple participation in the tab Questionnaire URL of the mailing.
• You can also create links yourself that carry a marker and send them to the respondents: Marking a URL. Please note that in this case the pseudonym appears both in the invitation and in the dataset.

Search terms: panel, panel studies, experience sampling, ambulatory assessment