The Roomba Paper

OK, I'll give you some supplemental information about the Roomba Study that isn't available on other links that will help you with your paper.  I'll also tell you exactly what statistics to run. 

You were observing  randomly selected mature Roombas (indeterminate sex - they can't be sexed visually - males and females are comparably sized) in their natural environment (a 4 x 6 ft black laboratory table).  Each observation on each tape may be a different Roomba, but you can't be sure, because the selection is "random."  Given that the herd of Roombas was estimated to be 500-800, we can be fairly confident that the majority of our observations were of separate individuals.  The tapes were made over a period of several weeks for each tape, see below.  The Roomba environment was marked off into quadrants and numbered using Liquid Paper, a substance that other unpublished observations had shown to be undetectable to Roombas. 

The Roombas were observed at the beginning of their "days" - they are diurnal animals.  In two of the tapes we provided supplemental feedings to the Roombas either on the left or right side of their environment.  From previous unpublished observations, we determined that Roombas readily consumed the 6.2 mm white paper disks we provided.  These disks were produced using a standard hole punch and Boise X-9 20 lb paper.  The disks were provided throughout the taping sessions to ensure a ready supply of this high density food.  In the third tape, there was no intervention, only naturally occurring foods could be consumed.

The reader will want to know how many observations were made.  The reader will want to know about how many observer pairs there were.  You need to report interobserver reliability data.  Toward that end, run X2 tests on the two observer's ratings of Behavior and Quadrant.  In each case, since there are more than 2 levels of each variable, you should check Contingency Coefficient.  Interpret the resulting values as you have been instructed to in the lecture portion of the course. 

Ignoring the Tape variable, report the frequency with which each Behavior occurs.  Because the frequencies of many behaviors are relatively low, we will collapse across behaviors and focus on whether the animal went straight (Y/N).  Of course, the other option that isn't straight is to turn in some way.  Now both observers may not agree on this new variable.  In your hypothesis tests, you were allowed to simply use either observer's data.  Here, we'll be more restrictive and create yet another variable that is Straight (Y/N) only IF both observers agree on the behavior.  The reader will want to know how many observation have been dropped to get to this point.  It relates to reliability and the issue of whether you have enough remaining data to tell a rational story. 

Now, to look for evidence of satiety, we will do X2s - no surprise!  First, we'll look for the frequency of being on the left versus right (sort of) while collectively considering only the two tapes with supplemental feedings.  We'll create yet another variable that isn't Right/Left, but will take those measures and convert them to Same/Opposite side of the Roomba environment relative to where the supplemental food is.  Additionally, since the Roomba should start out hungry and change its behavior after feeding, we will compare the 1st and 2nd halves of the observation period to create a 2 level Time variable.  So you will have a 2 X 2 X2 with the variables being Time and Side.  To get a clean look at the Same/Opposite concept, we'll do a straight X2 on only that measure as well.

Finally, we'll do a X2 on the Straight behavior while again collapsing the data from the two "fed" tapes into one level of a two level Fed (Y/N) variable.  Obviously, the unfed tape will be the other level of this variable.  Again we want evidence of satiety, so we will use the 1st and 2nd halves of the observation period for the Time variable.  So, we're going to end up looking at Fed X Straight twice - once in the 1st half and once in the 2nd half.  Obviously, if there is satiety in evidence here, there should be a difference between halves in the relationship between Fed and Straight. 

The bad news is, that the results of this study aren't crystal clear, in some ways.  You have to report it like it is and do your best in the Discussion to make sense of it.

Good luck!!  (Supplemental syntax)





.