One of the most common types of "information retrieval"

processes is document-database searching. An experiment was conducted to investigate the variables that influence search performance in the Medline database and retrieval system (Journal of Information Science, Vol. 21, 1995).

Simple linear regression was used to model the fraction y of the set of potentially informative documents that are retrieved using Medline as a function of the number x of terms in the search query. based on a sample of n = 124 queries.The results arc summarized below:

jj = .202 + .135x t (for testing H,: p, = 0) = 4.98 Two-tailed p-value = .001

a. Is there sufficient evidence to indicate that x and y are linearly related? Test using cu = .01.

b. If appropriate. use the model to predict the fraction of documents retrieved for a search query with x = 3 terms.