Section II: Behavioral Observation Methods

Lecture Plan

I.  Introduction to Methodology
          This unit teaches about important aspects of how behavioral observation data
are structured  for  analysis. It deals with one simple idea, patterning, but it requires
thinking about the methods involved. "Patterns" are repetitions of events in time.
There are two ways of talking about repetitions in time:
            (a) the total number of events per unit of time, or base rate, and
            (b) the specific sequence  of lead-follow units (e.g., Stimulus
                  Response events, SR  units as they occurred over time.

        For example: How many times do you touch your face in 10 minutes?
                                 (Total number of face touches.)
                 How many of those face-touch times occur when I call on you?
                         Now  the unit is--
                                   [ Stimulus: Calls on;   Response: Face Touches]

        If I know your touch rate is 4 times per minute, whether or not I call on you,
      I can also ask whether you touch more or less (than EXPECTED), given
      that I called on you. In terms of probability:

          we would determine that your rate of face touches is, say .25. (If I secretly
      looked in on you, you are likely to touch your face 25% of the time.)
      That is your base rate of face touches. It is also the unconditional probabilty
      (it just happens, it is not conditional upon somehting else having happened first).
      Now I can also calculate how many of those face touches happened when
      I call on you.

            Possibilities: all of them occur ONLY when I call you,  100%, or
                                 50% occur when I call on you.
                                 3% occur, and so on, all the way to 0% occur when
                                 I call on you.

                             So  you see,  the pattern  [calls on -- face touches] may
                            be more or less likely than just knowing  you are a 25% face-toucher.
                            But don't be confused: your total face-touch "pie" is still 25%.
                            We are only asking whether the events represented by the 25% of
                            the time occur more than expected ( i.e, just  25%)  when something
                            else has just  happened.

            That is, the conditional probabilit is that  probability of  some event,
                      given some other prior (conditional) event.

            For example, H+ given W- is NOT the same as the unconditional
            probability of H+. He may be  "more"likely to be  positive when she is
            negative than we expected, i.e., more positive than  just knowing his
            base rate of being positive (H+).

II.  Understanding methods for describing patterns of interactions

           Data stream = the actual depiction of the events according to the code-book.
             Base rates = the observed rate at which something happens over time,
                                 as in total events  divided by total time

                  E.g., 33 positives in 10 minutes = 33/10 = 3.3 per minute

            Sequential analysis = method for describing unfolding patterns
                            Retains the sequence (or order) of events, i.e., which
                            came first, etc.

                   E.g., H+ W- H- W- H+ W- W+ ..................

                      Example:  Base rates (from this sequence):
                                     2 H+, 1 H-, 1 W+, 3 W-  (in x minutes)

              Sequential analysis (Lag-1):

Stimulus:: H+ H- W+ W- Total
H+       2       2
H-     1 1        2
W+             (1)*
W- 1 1 1        3
                                                                        * Series ends with W+ so there is no "response" to it; not counted in Total

                                                                    Discussion Topic :                          Sex

   Fun   H+  H- W+ W-
  H+       27
  H-   36 -22 23 
  W-   29 -32 34
                                                Discussion:                                                  Sex
     Fun   W+|H+  W-|H- H+|W+ H-|W-
  W-|H-   39   62
  H-|W-   54   46
                                         See "Further explanation"
III. Coding Illustrations

         A. Coding Options -- different ways of portraying methods for observing
               interaction data

     B. Data Stream -- options for clustering data (i.e., discrete code, sequences)


     C.  Examples of cumulative point graphs

     These show the trend within a discussion,whether increasingly positive
            (regulated) or  increasingly negative (non regulated).  Using the trend
            (point graphs) as a predictor we  can then decide whether wehave types
            of couples (i.e., satisfied or  divorce prone, etc.).  Positive and negative
            codes are assigned numerical weights, and the algebraic sum of the
            weights are plotted for every talk turn. E.g., +5 -3 +2-1 =  +3. If the
            discussion gets more  and positive (i.e., +'s outnumber -'s) the trend
            is upward; scores are added to the immediately  previous score
           (= cumulative points).

                Continue to next page
       Behavioral Observation
 Return to Main