Knowledge Base

Lag Sequential Analysis in Detail

State Lag in Detail

A positive value of lag order N means that the program counts N events forward in time from the criterion event. The transition is established from the criterion to the Nth event. A negative value for N means that the program counts events back in time.

A state lag with order +1 results in the following transitions:

• X0 > Stand
• Stand > Sit
• Sit > Stand (2nd)
• Stand (2nd) > Walk
• Walk > Other locomotion
• Other locomotion > Y0

There are also transitions when a criterion is found but no target, or the other way round. In the example above, from no event (X0) to 0.0 Stand, and from Other locomotion to no event (Y0). See Understanding X0 and Y0.

A state lag with order -3 results in the following transitions:

• Stand (2nd) > X0
• Walk > Stand (1st)
• Other locomotion > Sit
• Y0 > Stand (2nd)

Gaps between events are not taken into account, so there is no transition from or to a gap.

Time Lag in Detail

The time lag can be negative, positive, or partly negative and partly positive (for example, from -2s to +2s around each event). Transitions are counted from an event to another event starting within the time lag.

Entering From 0 sec to 1 minute next to Time Lag results in the following transitions:

• X0 > Stand
• X0 > Sit
• X0 > Stand (2nd)
• Stand > Sit
• Stand > Stand (2nd)
• Sit > Stand (2nd)
• Sit > Y0
• Stand (2nd) > Y0
• Walk > Other locomotion
• Walk > Y0
• Other locomotion > Y0

Time Lag and Gaps

Transitions are also counted when no event is found in the time lag, for example at the start and end of the observation. See Understanding X0 and Y0.

If there is a gap between events, the program does not count transitions from an event to the gap that falls in its time lag, unless there is no event at all in that lag. When the time lag is partly negative and partly positive, transitions are calculated from the focal event to the events that precede or follow it.

To Ignore Recurring Criteria

If you select the Ignore Recurring Criteria option, all events of the same type as the criterion that occur within the time lag are ignored as criterion events. Select this option when events of the same type occur next to each other, and you want to make sure that transitions from those events are not counted twice.

If the time lag is at least partly negative, the recurring criteria are ignored within a time interval around the focal criterion whose length is twice as long as the longest side of the time lag. For example, if the time lag is set to from -10s to +5s, selecting Ignore Recurring Criteria removes the recurring criteria from -10s to +10s for each focal criterion. If the time lag is set to from -3s to +5s, recurring criteria are removed from -5s to +5s around each focal criterion.

To Ignore Recurring Targets

If you select the Ignore Recurring Targets option, the second, third, and subsequent instances of the same target within the time lag are ignored. Select this option when events of the same type occur next to each other, and you want to make sure that multiple transitions to those events are not counted.

If the time lag is at least partly negative, the recurring targets are also removed in the negative part of the lag.

Transitions with Events with Duration

Transitions indicate a change from one event (criterion) to another (target) in a temporal sequence. In the case of state behaviors, which have a start and a stop event, transitions are always counted using start events only, even if you have coded the end event manually.

Transitions and Instantaneous Sampling

If you have collected data with Instantaneous sampling, samples not scored are not taken into account when calculating transitions. For example, the sequence of samples (sample interval = 10s):

• 0.0 A
• 10.0 - (not scored)
• 20.0 B

Results in a transition from 0.0 A to 20.0 B. This is the same result as if A and B were scored with Continuous sampling.

Transitions and Multiple Event Logs

Lag sequential analysis is performed on each event log separately. If an observation contains multiple event logs and you de-select Event Logs in the Layout page of the Lag Sequential Analysis window, transitions are counted within each event log and then summed up in the transition matrix. Transitions across event logs are not calculated.

To calculate transitions from an event in one event log to an event in another event log, follow the procedure below:

1. In your data profile, select Intervals by Manual Selection.
2. Under Begin Interval, specify the event you want to use as a criterion (for example, Game starts).
3. Under End Interval, specify the end of the same event (if you want to analyze transitions as long as the event occurs) or a time value (if you want to apply a time lag; do this if Game starts is an event without duration).
4. Calculate statistics with this data profile. Make sure that you include the Total Number as a statistic. See Statistics in Detail in Calculate Statistics.

The Total Number column shows the number of times that a behavior occurs in the specified intervals. The Happy column/row displays the number of Happy occurrences in the specified time after Game starts took place. This provides results similar to time lag sequential analysis.

Source: The Observer XT 17 Help - Lag Sequential Analysis, Noldus Information Technology