NAME:   films dataset 
TYPE:   simple random sample 
SIZE:   100 observations, 5 variables  

DESCRIPTIVE ABSTRACT: Title, year of release, length in minutes, number of cast members listed, rating, and number of lines 
of description are recorded for a simple random sample of 100 movies.  One can use the sample to obtain base-line information 
on the movie guide from which the data were collected.  The dataset also illustrates two paradoxes for associations between 
three variables:  non-transitivity of positive correlation and Simpson's paradox.  SOURCE: The data were taken as a simple 
random sample of the approximately 19,000 movies (not including made-for-TV movies) in Leonard Maltin's Movie and Video 
Guide, 1996.  

VARIABLE DESCRIPTIONS:   
Columns  Variable   
1 -  35  Movie title  
37 -  40  Year (release year of the movie)  
42 -  44  Length (running time of movie in minutes)  
48 -  49  Cast (number of cast members listed in the Guide)  
54 -  56  Rating (movie rating on a scale 1, 1.5, 2, 2.5, 3, 3.5, 4)  
66 -  67  Description (number of lines of text to describe the movie)    

SPECIAL NOTES: The movies were obtained by drawing 120 pairs of the form (page number, item on page) to locate random page 
numbers and item number within a page.  A number M, no less than the maximum item number on a page, was estimated by skimming 
the Guide, then "item on page selections" were uniform from the integers 1, 2, ..., M.  The "page number" was a uniform 
selection from the pages in the Guide.  Maltin rates on a scale of 4 *s for the "very best", down to 1.5 *s for the "very 
worst", in half-step increments, with a special rating of BOMB for even worse than "very worst" movies.  I denoted these 
last movies with a number 1, and went from there to 1.5, 2, ..., 4.              

PEDAGOGICAL NOTES: These data can be used to teach basic EDA, confidence intervals for means or proportions, and relationships 
between quantitative variables.  The dataset illustrates both the non-transitivity of positive correlations and Simpson's 
paradox.  

REFERENCES:  
Langford, Eric, Neil Schwertman, and Margaret Owens (2001), "Is the Property of Being Positively Correlated Transitive?" 
_The American Statistician_ 55, 322-325.  

Maltin, Leonard, _Leonard Maltin's 1996 Movie and Video Guide_, Penguin Books, NY, 1996.   

SUBMITTED BY: 
Thomas L. Moore 
Department of Mathematics and Computer Science 
Grinnell College 
Grinnell, Iowa 50112 
mooret@grinnell.edu