Observing and Quantifying Behavior – Lab 2



Observing and quantifying behavior

A 2-3 hour laboratory exercise

Modified from

© 2002 Daniel T. Blumstein & Janice C Daniel

Today, you will be working with different recording methods that may be useful for field projects or future studies. Specifically, you will learn how to quantify behavior using both time and continuous recording for focal sampling of subjects. You will also learn a few additional methods to quantify your inter-observer reliability.

We will use several clips of flamingos videotaped at the LA Zoo as an example of how to calculate time budgets for animals. Time budgets describe how animals allocate their time to different behaviors. The calculation of time budgets can be used to show how an animal’s behavior varies with changing circumstances, such as predation risk, group size, seasonal changes, etc.

Because flamingos generally live in groups, we could potentially estimate their time budgets using either scan sampling (rapidly scanning a group of subjects) or focal sampling (focusing on one subject at a time). These two methods yield somewhat different results. For the purpose of this exercise, we will use focal samples. You will score behavior from three different individuals, each tracked separately. Each focal sample is approximately 2 minutes in duration.

What are the advantages and disadvantages of using scan sampling compared to focal sampling for calculating time budgets of these flamingos? In general?

PART 1 – Time Recording

BEFORE WE BEGIN – Log onto the computers as yourself. Log onto the Courses Server as yourself.

First, we will begin with time recording (recording behavior at a specified interval of time). We will sample the behavior of a focal individual every 10 seconds. After scoring all three clips, you will calculate your inter-observer reliability (consistency or repeatability) with a partner.

In general, time recording is an efficient and often easy way of collecting data in the field that does not necessarily require any “high-tech” equipment. Pencil, paper, and a simple timing device (such as the stopwatches you used during the Planarian learning lab) are all you need, although more fancy event recorders may also be used. Later today, in lab, you will use the JWatcher event recorder to collect data.

A) Viewing the ethogram

An ethogram is a catalogue of behaviors described by neutral postures and recognizable movements. The specific ethogram used in any study will depend upon the question that is being asked, and also on how accurately, and with what specificity, a particular behavior can be scored. We will familiarize ourselves with the behaviors shown in this flamingo ethogram by watching the video.

The Ethogram is contained within the QuickTime Player file called “Ethogram_Comp.mov.” that Ned (Carey) will play a few times for the class.

Note that this ethogram has been chosen arbitrarily.

What potential problems do you foresee, if any?

Why is it important to have a category for “out of site”?

Why is it important to have a behavior category for “other”?

B) Collecting data using Time recording of a focal sample

Look at the data sheet (Time Recording Data Sheet) at the back of this exercise. You will see the behaviors from the ethogram listed across the top of the page. You will also see a series of rows. Each row will correspond to a tone that you will hear every 10 seconds when viewing the subsequent three focal clips. You should make one check mark per row, in the column corresponding to the behavior that you observe at the tone. Notice at each tone (time point) you will check one and only one behavior. For today, all behavior are considered to be mutually exclusive. Two behaviors cannot happen at the same time. JWatcher can do multiple and modified recording, but that is beyond the scope of today’s project. Even if a very interesting behavior happened between the beeps you will not check that behavior. Your score sheet should be an accurate representation of what the animal was doing at the instant of the tone.

There are three video clips. The class will watch each video clip once. Since you are recording manually, each student should record his or her own data. You may use the score sheet provided and attach that to your lab notebook or record directly into your lab notebook. In either case, do not forget to record which movie you are watching.

Ned (Carey) will first open and play “FlamingoA_Comp.mov”

There is a 5 second delay before the video begins. The individual that is chosen to be the focal animal is labeled in the video with an arrow. Before the movie starts, be sure that everyone agrees on which animal is the focal animal.

Record the behavior that you observe at each tone in your data sheet.

Ned (Carey) will next open and play “FlamingoB_Comp.mov”

Record the behavior that you observe at each tone in your data sheet.

Finally, Ned (Carey) will open and play “FlamingoC_Comp.mov.”

Record the behavior that you observe at each tone in your data sheet.

C) Calculating reliability

When more than one researcher is collecting data for a project it is necessary to check that everyone is consistent in their recording.

1. On your data sheet, add up the number of times you observed each behavior for all three video clips. Enter the totals in the 2nd to last row of your data sheet.

2. Ask someone to be your partner so that you can compare your results.

3. Discuss your results. What behaviors were most accurately scored? What types of behaviors were more controversial?

Although eyeballing totals is one way to get a feel for your consistency, there are more formal ways of calculating reliability. We will use an Excel spreadsheet (something new, not statview) to calculate three measures: r, O, K. (method O is the method we used in the Planarian Learning lab)

r is the correlation coefficient calculated between your observed totals for each behavior and your partner’s corresponding totals. An r value near 1 means that you were reliable.

O is the index of concordance which is simply the proportion of observations on which you both agreed (scored the same behavior at the same tone.) It is calculated by counting the total number of agreements and then dividing by the total number of observations.

K is the kappa coefficient, which is basically an index of concordance that has been modified to take into account the number of agreements that you would get simply due to chance alone.

4. On the Courses Server go to:

Biology/Bio101 102/Behavior_Renn/Week3_flamingos/ and in that folder your will find an Excel file called reliability_template.xls.

5. Drag reliability_template.xls onto your desktop.

6. Double click on this icon to open the file.

7. Enter your observations, coded by number into the O1 column (for observer 1).

8. Have your partner put their observations into the O2 column (for observer 2). Notice that you enter the number that corresponds to each behavior (see top row of your check sheet) rather than letters. What type of data is this, categorical (nominal) or ordinal (continuous)?

9. The spreadsheet will calculate r, O, K for you automatically.

10. Print this record for your lab notebook.

• How reliable was your scoring?

• When might it be important to check O rather than just r?

• Which behaviors were consistently scored? Which were not? Can you explain the inconsistencies? How might you improve your reliability in the future?

• How would you measure your intra-observer reliability?

D) Calculating the time budget

Finally, in your check sheet, or your lab notebook, calculate the proportion of time engaged in each behavior by dividing each column total by the total number of observations. This is your time budget.

Do you think that a ten second time interval enabled us to adequately describe the time budget of these flamingos? Why or why not? In general, what criteria should you use to choose an appropriate time interval?

PART 2 – Continuous Recording

Now we will go back and score the exact same three video clips using continuous recording with an event recorder. Event recorders are computer programs (or dedicated pieces of hardware) that record keystrokes as they occur over time. In our case, keystrokes will represent behavioral transitions. For instance, when a flamingo is foraging, you might type an f, when a flamingo is grooming, you might type g, etc.

Continuous recording (all-occurrence recording to provide a complete record of all behaviors) is more labor intensive, but will give you a more complete description than time recording of what occurred during any given time interval.

We will use an event recorder called JWatcher to score our three video clips continuously The JWatcher program includes some analysis algorithms along with the power of the event recorder. With these tools, you will be able to calculate the time allocated to each of the behaviors in your ethogram. This will give you a more accurate time budget for zoo flamingos. The full JWatcher program and manual is freely available on-line at: jwatcher.ucla.edu.

This flowchart describes JWatcher work flow.

By the end of lab today you will:

A) Create a Global Definition File (*.gdf) which will serve as the “ethogram” and assign keys to each behavior.

B) Create a Focal Master File (*.fmf)

which includes the behaviors (or a subset thereof) from the global definition file. This file also instructs JWatcher to prompt users for additional data about each observation during data collection.

C) Record behavior. The data file (*.dat) file is created automatically during each observation. It contains the additional data about the observation as well as the raw data that you have recorded

D) Create a Focal Analysis File which specifies how data files (.dat) will be analyzed and specifies the results to be reported.

E) Quantitative analysis of each *.dat file will produce a results file (*.cd.res) that can be opened in Excel.

JWatcher is capable of more complex observation recording (behavior combinations) and also more complex analyses (transition and Summary results) that we will not explore today.

A) Defining your ethogram in JWatcher

You will continue to work in pairs.

The first thing you need to do is “define” your ethogram in JWatcher. Each student pair will be recording the same behaviors we have recorded previously by hand, but you can adapt your ethogram key code to suite your own habits. Some people prefer to use logical key locations, others prefer to use logical letters that match the behavior description. Confer with your partner and decide on a code.

1. In the Applications folder, find the JWatcher_V1.0 folder and double click. Now double click on the JWatcher icon (black square, with a gold J and gray W) to launch the JWatcher program.

2. When JWatcher first opens, you will be in the Data Capture tabbed window. Click on the Global Definition tab to change windows. The Global Definition window is where you will specify your ethogram (by assigning key codes to each behavior.)

3. The Global Definition window is divided into two sections: one to define behaviors and another to define modifiers. We will be defining behaviors only.

4. Click on the Add row button in the top section to begin assigning key codes to behaviors.

5. It is up to you and your partner to choose the key strokes that correspond to each behavior. You may use only one single character per behavior. Try to choose something intuitive and easy to remember, such as a for aggression and f for forage, etc. Note that JWatcher is case sensitive: f will NOT be recorded the same way as F. Because it is difficult to switch between upper and lower case, for this exercise, use lower case letters. Record the letters that you choose for each behavior in your lab notebook also.

6. Type in the single character code in the Key Code box. Then either tab or click into the adjacent Behavior box to type in the corresponding behavior. You may also include a more lengthy description of each behavior in the Description box but remember to avoid anthropomorphisms or “intent” in your descriptions whenever possible.

7. Click on the Add row button to continue adding each behavior.

8. When you are done adding all your behaviors (you should have 10 behaviors), click the Save button in the lower right corner of the window.

9. To Save your new Global Definitions file to the desktop click the house icon in the upper right part of the Save As window. Now double click on the Desktop folder icon for local user. Name your file (gdf_yourname_Ned_tues) by typing into the File Name: box in the lower part of the window. The extension .gdf will automatically be added when you click Save (this stands for Global Definitions File).

Write down your chosen key codes and behavior descriptions in your lab notebook and record the key codes at the top row of the data sheet (Continuous Recording Data Sheet) included at the back of this exercise.

B) Specifying focal duration and other details for a JWatcher project

1. Now that you have specified your ethogram, the next step is to specify the length of time that you would like to collect data for each focal sample, and also what (if any) additional information you might want to record for each focal sample, such as the time of day, or individual observed, etc.

2. Click on the Focal Master tab at the upper left to change windows (be careful, as it is right next to Focal Analysis Master). (You may find yourself in the Test Details page (see bottom left tabs). If so, change pages by clicking on the Define Codes tab.)

3. The Define Codes page shows the ethogram to be used when scoring data for this particular project. Currently, it should be blank. You could either reenter your ethogram, or import the global definition file that you have already created.

4. To import your pre-existing ethogram, click on New (bottom center of the window) and a new selections window will open. You should see your .gdf file that you just made (gdf_yourname_Ned_tues.gdf) (if not, you will have to navigate back to the desktop by clicking the house icon and then the desktop icon). Select your global definition file (yourname_Ned_tues.gdf) from the list in the top part of the window. Click Open to load your ethogram.

5. Next, using the lower left tabs, change to the Test Details page. We will score each video for exactly 2 minutes. In the Duration boxes, type 0 in the hours (HH) box, 2 in the minutes (MM) box, and 0 in the seconds (SS) box.

6. Finally, in the lower left tabs, change pages again by clicking on Questions. This window is where you may define the variables that will be associated with each focal sample file, such as individual ID, age, sex, time of day, etc. For our purposes today, you only need to record the video clip name (A, B or C) for each sample. But think about what types of information you would want for a real study.

7. Type in Video Clip Title? under Question 1. This will prompt you later to specify which clip you are watching when you start each behavior recording.

8. Click Save As to save your Focal Master File (fmf_yourname_Ned_tues). Make sure that this file is being saved to the desktop where you saved your .gdf file (though you will not currently see that file name). Again, the proper file extension for a Focal Master File (.fmf) will automatically be added for you when you click Save.

C) Scoring behavior with JWatcher

Now you are ready to begin scoring behavior.

Change windows by clicking on the Data Capture tab (upper left).

1. In this window, click on the file navigator icon (it looks like a sheet of paper) to the right of the Focal Data box in order to name the data file you will be creating. Navigate to the desktop by clicking the house icon then double click the desktop icon.

2. Type in the name of your new file dat_name_Ned_tuesA in the File Name box and click Open. Keep the names simple. (use only one of your names, Jwatcher sometimes has problems with long names and spaces in names)

3. Specify your Focal Master File (fmf_yourname_ned_tues.fmf) by clicking on the file navigator icon (it looks like a sheet of paper). You may need to navigate to the desktop.

4. Select your .fmf file. Click Open.

(Remember that this file contains the ethogram along with additional specifications such as the duration for recording the focal observation.)

5. Click the Next button at the bottom right to tab into the next page.

6. Answer the top question by typing the letter of the video clip into the box below the question and click the next button to advance. We will start with A

7. Click the Next button.

If you wish to see a list of the behavioral codes on your computer screen, click the Behaviors box in the upper right corner of the JWatcher screen.

8. Wait for Ned (Carey) to start the video clip. You also have to wait for the class.

9. You should be looking at the JWatcher – Data Capture – data_yourfile.dat window

10. When the film has started, click on the Start button in the center of the Data Capture page.

11. Immediately type the key code representing the behavior the subject is currently engaged in.

12. Whenever the behavior changes, type the key code for the new behavior. Continue for the two full minutes.

(JWatcher will automatically time out).

(JWatcher will automatically save your data in the .dat file that you named above)

If you made a data entry mistake, look at the timer and make a mental note of it. You can fix it later. Because the whole class is working in parallel, we cannot backup and watch the videos over.

When the class is ready to start the next video:

1. click the New button in the lower right corner.

2. Enter a new name for the next data file dat_name_Ned_tuesB in the File Name: box.

3. Click Open.

4. Do not change the focal master file name. Leave it set to fmf_yourname_ned_thus.fmf

5. Click Open. Click Next. Answer the question. Click Next.

6. When the film starts Click Start.

REPEAT THIS FOR THE FINAL VIDEO

D) Using JWatcher to analyze focal animal samples

Once you have scored all three video clips, you are now ready to use JWatchers analysis algorithms to calculate time budgets. Analyzing data files is a two-part process.

First, you must create a new file called a focal analysis master file (*.faf) that specifies which analyses you want done.

Second, you must run your data files through an analysis routine to generate results files.

1. To create a focal analysis master file, click on the Focal Analysis Master tab in the upper left menu to open the window.

2. Click on the New box, at the bottom center.

3. You must now load your projects ethogram by choosing the same focal master file that you used to capture data (fmf_yourname_ned_tues.fmf.) (If you do not see your *.fmf file currently displayed, then navigate to the desktop by clicking the house icon and then the desktop icon.) Select your file and click Open.

4. Once your ethogram has been loaded, immediately change pages by clicking on the State Analysis tab on the bottom series of tabs. (Note: there is also an Analysis tab above and a States tab below, you want State Analysis)

5. This is where you will specify which type of state analyses you want performed. For this exercise, we will use the All Duration column on the far right. This means that we are interested in the duration of states (but not the intervals in between), and also that we want to include information about all of the states in our focal sample, regardless if they have been truncated by the beginning or ending of the session.

6. Select all the options available in the All Duration column (you should check a total of seven boxes). Also, to the right, designate your key code for out-of-sight. If your animal was out-of-sight for some part of the focal, then JWatcher will correct for this by calculating the proportion of time in sight.

7. Click on the Save As button. Specify the location to save your new focal analysis file, and give it a name (faf_yourname_ned_tues) by typing into the File Name: box. Navigate to the desktop if necessary and click Save. Again, the proper file extension .faf will automatically be added.

8. To generate results files, click on the Analysis tab in the center upper menu.

9. Use the file navigator icon (looks like a piece of paper) to select your faf_yourname_ned_tues.faf file. Remember that the focal analysis file specifies the types of analyses were going to calculate. Click Open.

10. Use the file navigator icon (looks like a piece of paper) to select the data file (data_yourname_ned_tues.dat) to select the data file to analyze. Click Open.

11. Specify the Desktop as the appropriate location as the destination for the results by clicking on the folder icon and then click Open (note: this file is automatically created and named, you are just verifying that you know where it will be placed).

12. Select the Print results for all behaviors radio button.

13. Click the Analyze button at the bottom of the file window to analyze your data.

14. If you get a warning message, this is simply to tell you that there were unrecognized key strokes in your file. These will be ignored.

REPEAT the analysis for each of the other two data files.

dat_name_Ned_tuesB and dat_name_Ned_tuesC

E) Viewing the results files

You will use Excel to view the results files.

There are two results files, *.cd.res, and *.tr.res.

The *.cd.res file has the quantitative results.

The *.tr.res has a file that you can open and graph the behavioral traces.

For todays exercise, you will open the *.cd.res files.

View the *.cd.res file by opening it through Excel. Excel will not automatically recognize JWatcher files; you will need to tell Excel do it. The *.cd.res file is a comma delimited text file.

Launch Excel from the dock

Use File /Open and navigate to the desk top.

At the top center of the Open dialog box Enable: All Documents needs to be set.

Select dat_yourname_Ned_tuesA.cd.res (your analyzed results file)

Click Open

Select the Delimited radio box

Click Next

Check the Comma box and Uncheck the Tab box

Click Next

Click Finish

Once opened, you should see a list of the behavioral codes and several summary statistics for each behavioral code. Depending on the focus of your research you may be interested in different measurements of the behavior. Today we will use the number (N) of times a behavior occurred and the proportion (Prop_IS) of the total time in sight that is accounted for by each behavior.

Each of your result files will have the following column headings that refer to just one of your video observations.

|StateAllDur_N: |Occurrence |

|StateAllDur_TT: |Total Time (in milliseconds) |

|StateAllDur_X: |Average Time (in milliseconds) |

|StateAllDur_SD: |Standard Deviation of the average time (in milliseconds) |

|StateAllDur_PROP: |Proportion of Time |

|StateAllDur_PROP_IS: |Proportion of Time in Sight |

If your animal never went out-of-sight, then proportion of time and proportion of time in sight should be the same. The proportion of time in sight for your behaviors is your time budget.

Open your other two results files.

If you want to print these out, please save paper and take the time to format in Excel so that it prints the complete file on one page. Select the row with the column headers. Format / Cell / Alignment check wrap text.

Decrease the column width of all columns to fit on one page. (Ask if you need help.)

F) Comparing continuous recording versus time recording

Use the “Continuous Recording Data Sheet” at the back of this exercise or in your lab notebook to compile your JWatcher analyzed data for these three observations.

We are going to compare the data collected by our JWatcher continuous recording technique to the data collected by our manual time recording technique.

1. Fill in the N for each behavior in each video.

2. Count the total N for each behavior across the three videos.

3. Fill in the Prop IS value for each behavior in each video.

4. Determine the average proportion of time.

5. Compare the two Total counts for each behavior that you recorded using the two techniques. Use the rough grid provided on the last page of this handout to make a bar graph representing the total events recorded with time and continuous recording. Add the proper scale markers to the graph, and label the X axis behaviors

6. Compare the two Time Budgets that you calculated using the two techniques. Use the rough grid provided on the last page of this handout to make a bar graph that represents, for each behavior, the percent of total behavior recorded with time recording and the average proportion of in sight time that was recorded by continuous recording. Add the proper scale markers to the graph.

These rough graphs will give a visual representation allowing you to compare the two techniques easily. While these graphs are not highly accurate, they may be very useful in evaluating which technique to use in future studies.

Did continuous recording give you a better estimate of flamingo time budgets than time recording?

Were some behaviors captured equally well with the two techniques?

If you were to collect more data, which recording rule would you use in the future? Why?

Under what conditions would you choose not to use an event recorder?

How would you test for the reliability of your continuous recording technique?

Time Recording Data Sheet

|1 |2 |3 |4 |5 |6 |7 |8 |9 |10 | |Tone |Aggress |Forage |Groom |Head

Swing |Stand Look |Sleep |Walk |Wing

Spread |Other |Out

Sight | |A.1 | | | | | | | | | | | |A.2 | | | | | | | | | | | |A.3 | | | | | | | | | | | |A.4 | | | | | | | | | | | |A.5 | | | | | | | | | | | |A.6 | | | | | | | | | | | |A.7 | | | | | | | | | | | |A.8 | | | | | | | | | | | |A.9 | | | | | | | | | | | |A.10 | | | | | | | | | | | |A.11 | | | | | | | | | | | |A.12 | | | | | | | | | | | |A.13 | | | | | | | | | | | |B.1 | | | | | | | | | | | |B.2 | | | | | | | | | | | |B.3 | | | | | | | | | | | |B.4 | | | | | | | | | | | |B.5 | | | | | | | | | | | |B.6 | | | | | | | | | | | |B.7 | | | | | | | | | | | |B.8 | | | | | | | | | | | |B.9 | | | | | | | | | | | |B.10 | | | | | | | | | | | |B.11 | | | | | | | | | | | |B.12 | | | | | | | | | | | |B.13 | | | | | | | | | | | |C.1 | | | | | | | | | | | |C.2 | | | | | | | | | | | |C.3 | | | | | | | | | | | |C.4 | | | | | | | | | | | |C.5 | | | | | | | | | | | |C.6 | | | | | | | | | | | |C.7 | | | | | | | | | | | |C.8 | | | | | | | | | | | |C.9 | | | | | | | | | | | |C.10 | | | | | | | | | | | |C.11 | | | | | | | | | | | |C.12 | | | | | | | | | | | |C.13 | | | | | | | | | | | |Total | | | | | | | | | | | |Percent | | | | | | | | | | | |

Continuous Recording Data Sheet

|Key Code: | | | | | | | | | | | | | |Agress |Forage |Groom |Head Swing |Stand Look |Sleep |Walk |Wing Spread |Other |Out Sight | |N |A | | | | | | | | | | | | |B | | | | | | | | | | | | |C | | | | | | | | | | | | |total | | | | | | | | | | | |

PropIS |A | | | | | | | | | | | | |B | | | | | | | | | | | | |C | | | | | | | | | | | | |average | | | | | | | | | | | |

This text was modified from Learning the scientific process by observing and quantifying behavior© 2001-2006, Daniel T. Blumstein and Janice C. Daniel All rights reserved.

It was downlaoded with permission from 20070306

It has been modified for educational uses only.

[pic]

Figure 1. Number of events recorded with Continuous Recording and Timed Recording. Total number from three 2 minute Flamingo videos.

[pic]

[pic]

Figure 2. Time Budget as calculated by Continuous Recording and Timed Recording. Continuous Recording as performed by JWatcher analysis for Proportion of time in sight. Timed Recording is reported as percent of total behaviors. Data sample is as in figure 1.

................
................

In order to avoid copyright disputes, this page is only a partial summary.

Google Online Preview   Download