Comparison the Observer vs ELAN

Last modified by Iris Spruit on 2024/12/02 12:07

Contents

	The Observer	ELAN
Website	www.noldus.com	https://tla.mpi.nl/tools/tla-tools/elan/
Price	~€3000 per basic license, per year	Free
Short description	The Observer is created for coding behavior in video’s.	ELAN is created for annotating (coding) and transcribing video and/or audio. It seems more focused on transcribing audio/video, but can also be used for coding video’s.
Info	The Observer has a useful video tutorial. After opening the Observer, go to Help à Video Tutorial.	Official documentation can be found here: Documentation \| The Language Archive (mpi.nl). Additionally, these video tutorials are highly recommended: ELAN Intro - ELAN explained in plain LEGO - VILA VIDEO TUTORIALS (youtube.com).
Project setup	One project is created which consists of: Setup: Observation method Independent variables Coding scheme Observations Analyses	One project per observation needs to be created. A template can be made of the settings and tiers that are created, which can be applied to further observations. Important settings in the template are: Tiers Tier types Controlled Vocabularies
Observation methods	The Observer distinguishes between three observation methods that are set for each project: Continuous Sampling: You record all occurrences of the behaviors of interest of one or more subjects as they occur. You obtain full descriptive statistics of the behaviors, like the duration per instance, the total duration and frequency. Choose Continuous Sampling to calculate absolute frequencies and durations of the behaviors of interest. This method is comparable to event sampling. Instantaneous Sampling: You record the behavior of one or more subjects at preselected moments in time. You obtain frequencies of the behaviors, not their durations. This method is comparable to time sampling. Combined Sampling. A combination of Continuous and Instantaneous Sampling.	A method cannot be set. ELAN is best suited for a continuous sampling method, however, instantaneous sampling would be possible with some workarounds.
Independent variables	Yes, independent variables can be added to the project and filled in per observation.	No
Live observation	Yes	No
Multiple media synchronization	Yes Multiple media (video) files can be added to an observation and synchronized with the Offset button in the relevant Observation.	Yes Multiple media (video) files can be opened when starting a new project. The media files can be synchronized in Media Synchronization Mode.
Coding/ annotating	Quick, easy and flexible The Observer works with Keycodes to code behavior fast with keyboard shortcuts, which are created by a coding scheme. In the coding scheme the following can be defined: Subject Behavior. Behaviors can have a duration (state events) or not (point events). Options: Mutually exclusive: A group in which behaviors cannot occur at the same time. Exhaustive: Always one behavior is active. Start-Stop: Behaviors can overlap and there can be gaps between behaviors, therefore the start and stop of each behavior needs to be coded. Modifiers: An extra property of a behavior (e.g. play with 1) ball, 2) cards, 3) castle, or intensity of aggression 1 – 5).	Can be a bit more complex and time consuming than The Observer, but once understood, quite easy and straightforward. All annotations (behaviors) have a duration (what the Observer calls state events). ELAN works with Tiers. By convention, each tier has annotations of the same type of events (e.g. hand gestures participant A, facial expressions participant B). Annotations on the same tier cannot overlap. Each Tier should have a Tier Type, which need to be created. A Tier Type is for example Hand Gestures. A Tier Type can have a Controlled Vocabulary, which are predefined codes for certain behavior that is coded in the video, which is useful when only certain known actions are coded. It is not possible to create custom keycodes to quickly code behavior, like in the Observer. However, keyboard shortcuts exist for viewing the video and creating annotations. Additionally, To quickly code video’s, use the Segmentation mode, which allows you to use the keyboard for creating annotations but without naming them. Afterwards, use Annotation mode to name the annotations, which is done quickly when a Controlled Vocabulary is used. This method should allow for a more quick and easy way of coding video’s.
Video playback controls	Video playback can be controlled: Speed up/down Quick review to jump back few seconds Loop to view short period multiple times Keyboard shortcuts for video playback exist	Video playback can be controlled: Speed up/down Jump back or forward 1 section, 1 sec, 1 frame, 1 pixel Loop to view short period multiple times Keyboard shortcuts for video playback exist
Analysis: statistics	The Observer has many options when it comes to visualizing, organizing and filtering the results. This makes analyzing data with the Observer not straight forward. The following statistics can be obtained per observation or over all observations; and per behavior (annotation). Mean duration Total duration Total number Minimum duration Maximum duration Standard deviation Standard error Rate per minute Percentage Latency 25th percentile Median 75th percentile The layout of the output table can be customized.	Simple annotation statistics can be obtained. The following statistics can be obtained per Annotation, Tier, Tier Type, Participant, Annotator or Language: Number/Occurrences Minimal duration Maximal duration Average duration Median duration Total annotation duration Annotation duration Percentage Latency
Analysis: filtering	The data can be filtered before doing analyses: Analyze a selection of behaviors, subjects or observations. Analyze clusters of behavior. Analyze a selection based on a minimum duration of an event.	Filtering can be achieved by creating custom Tiers. Clusters of behavior can be created in a new Tier by going to Tier > Create Annotations from Overlaps. This option can be applied to multiple files by going to File > Multiple File Processing > Annotations from Overlaps. It does not seem possible to filter behavior (or create a new Tier with filtered behavior), based on the duration of behavior. This could be achieved by exporting the raw data and then applying filters to it outside of ELAN (e.g. R).
Analysis: lag sequential	A lag sequential analysis can be performed in the Observer.	ELAN does not feature a lag sequential analysis option. Instead, lag sequential analyses can be performed outside of ELAN in R, see LagSequential: Lag-Sequential Categorical Data Analysis.
Analysis: reliability	A reliability analysis can be performed in the Observer.	A reliability analysis can be performed by going to File > Multiple File Processing > Calculate Inter-Annotator Reliability.
Result export format	Results can be exported to an Excel file.	Statistics can be saved in a text file. As an ELAN project features only one observation, one text file per observation is saved. Data of the observations need to be combined outside ELAN.
Import events	Events can be imported into an observation. Some standard types of event files can be imported, such as Tobii Studio .tsv files or an observer data file .odx. It is also possible to import a custom event file. The event file should be a .txt file. To import a custom event text file, a custom import profile needs to be created, in which you specify what the text file looks like and what the time and event columns are. The custom event text file should at have: A header A column with timestamps (in ms, sec, min, hours, days or a date/time format) On or more columns with event information: Behavior (event) Subject Modifier Comment Event Type The data column delimiter can be tab, comma, space, semicolon, or user defined.	Events can be imported as annotations. Some standard types of file can be imported, such as a PRAAT file. It is also possible to import a custom event file. The event file should be a .txt, .csv or .tsv file, with comma or tab separated columns. The file can contain header rows, which can be skipped when the file is imported. The custom event file can have the following columns: Annotation (event) Tier Begin Time End Time Duration The Begin Time, End Time, and Duration can be formatted as hh:mm:ss.ms or s.ms When the event file is not synchronized with the video file, one can do so by Playing the video and determining the moment where the Event file is timepoint 0. Going to Edit > Linked Files Fill in the offset in the video This method needs the following requirements: The video needs to be started before the event file’s timepoint 0. It needs to be visible in the video where timepoint 0 of the event file is. For example, when events are sent from stimulus presentation software, the video could register the start of the task.
Verdict	The Observer was designed for video coding. Its setup and usage is generally easy and user friendly. Using keycodes to code video’s is easy. There are many analysis filtering options. The software is however closed-source and an expensive license is required to access it.	ELAN was designed for video and audio annotation and transcription. Its setup is not as user-friendly as the Observer, however, most of what can be achieved with the Observer is possible in ELAN. For some, it may be more time consuming or at least an adjustment is necessary when setting up a project and coding video’s. Most analyses can be performed in ELAN, some need to be performed outside of ELAN, in R. A huge advantage of ELAN over the Observer is that it is free and open-source, thereby allowing researchers to always access their data in ELAN, without the need for a license. In sum, ELAN seems to be a very good alternative to the Observer.