In
the 1999 "Homelessness in Texas" statewide survey, data was collected
from 1720 homeless individuals, in 9 different cities. This large sample
allows us to hypothesize that the results found would be similar if all
the homeless individuals in Texas were interviewed.
However, the chances to make an error when generalizing a specific
result for the entire state are somewhat increased because the sample was
not randomly selected. For example, none of the 3 biggest cities of Texas
(Dallas, Houston and San Antonio) participated in the survey.
It is impossible to quantify how large is the probability of error.
It may be very small for some variables and large for others. For example,
the variables that are not influenced by city characteristics are very likely
to have the same distribution in all cities of Texas, and the hypothesis
tests based on these variables are very robust.