Agenda - 04-21-1998 - 10b - Laserfiche WebLink

Browse Search

15 primarily due to follow -up telephone calls made by the interviewers and project managers. Of these, 455 surveys were ultimately found to be usable in the analysis. Exhibit 2 is the disposition sheet used to track each survey. The completed telephone interviews were checked, and the information was carefully coded in order to stress quality control. The PI also examined every survey for coding errors and double checked questionable entries. Therefore, the information on which the analysis is.based is highly reliable. The data were compiled in a SAS data set and analyzed on a univariate and multivariate basis. The univariate analysis uncovered some questionable values in the data set, and again the original survey information was reviewed and corrected as appropriate to eliminate these discrepancies. Results The variable names and summary statistics are in Exhibit 3. The statistics were run on all variables for convenience although they are not meaningful for some variables such as the first four listed. The first 25 listed variables, through Interviewer Number, come from the compiled surveys. The next eight unlabeled fields convert the residential tenure variables from year -month values to decimal values in years. This conversion is accomplished by dividing "months of residence" by 12 to create a decimal value and then adding this amount to "years of residence." For example, the values for YRSRES and MNTRES are 3.24 years and 2.07 months, respectively. The decimal value, TOTRES, is 3.41 years which equals 3.24 years plus another 2.07/12 years. Approximate monthly rent was assigned to each apartment unit on the basis of its urban -area location and number of bedrooms. Assigning approximate rent level was preferable to asking respondents about rent they paid because survey questions about monetary values always drive down response rates to that question and to others as well. For example, 218 owner- occupants indicated the year their residence was built, but only 169 of them indicated its approximate value. The last two variables listed were computed from the data set. The number of adults is the difference between number of persons in the unit and number of children aged 18 or younger. The number of children in pre- school, private or home school is the difference between number of children and the number in public school in one of three grade levels.