You are here

Home

NCHS Research Data Center

NCHS Research Data Center

Slide 1

CDC/NCHS.Research Data Center
Presented July 23, 2003
Bureau of Transportation Statistics
Confidentiality Seminar Series

Kenneth W. Harris
Acting Director
(301) 458-4262
Kwh1@cdc.gov

Vijay Gambhir
Computer Scientist
(301) 458-4226
Vgambhir@cdc.gov

Slide 2
Operational Comparison of the NCHS Research Data Center and the Census Research Data Center

  NCHS RDC Census RDC
Data available Virtually any NCHS survey without direct identifiers Title 13 data such as CPS, MEPS, and others
Researcher-supplied data Allowed Allowed
Research proposals Required Required
Review cycle Continuous Three times per year
Turn-around time 2-3 weeks 4-8 months
Tenure at RDC Short term or long term (minimum charge of 2 days) Minimum 3 months
Remote Access Yes No

Slide 3

  NCHS RDC Census RDC
Type of project Tabular or model-based Model-based only
Researcher costs    
Remote access $500/month if file size < 130,000 records.
$1000/month if file size > 130,000 records		 N/A
On site $200/day  

Slide 4
Operational Comparison of the NCHS Research Data Center and the Census Research Data Center

  NCHS RDC Census RDC
File set up $500/days effort N/A
Computer equipment    
Hardware Windows NT server/Windows 2000 workstations Hardware: Unix serverAs described in Reznek talk
Software SAS/Fortran/StataOthers available upon request SAS/StataOthers available upon request

Note: At this time files containing restricted or confidential data cannot be transmitted across data center boundaries.

Slide 5
Analytic Data Research by Email (ANDRE)

  • NCHS has been providing remote data access to the researchers through ANDRE since April 1998
  • In the past five years ANDRE has served 45 different data analysts and executed over 10,000 SAS programs for their research programs.

Slide 6
Main Features of ANDRE

  • Completely automated system. Operates round the clock without any human intervention
  • Registered subscribers only.
    • Proposals already reviewed and approved
    • Have an agreement with NCHS/RDC
  • Unlimited access during the subscription period

Slide 7
Data Requests

  • Registered user can submit data requests by email from anywhere and at any time.
  • Results of the data request released to a specified email address that has been certified to be secure by the subscriber and approved by NCHS/RDC

Slide 8
Authentication

  • Multi-levels of system security:
    • Submission syntax
    • User id
    • Password
    • Email/code word
    • Package
    • Path info

Slide 9
Data Request Analysis

  • Compliance with the disclosure limitation constraints of NCHS
  • Integrity of the system
    • Resource constraints (CPU time & Storage requirements)
    • Protection of ANDREs work environment

Slide 10
Prevention of Direct Disclosure

  • Cleaning up of the Log File
  • Categorization of SAS commands/words
    • Forbidden Commands
    • Modifications to the Commands
    • Output suppression

Slide 11
Sample:- Original Log

1 options nocenter;
2 Data one;
3 Infile 'd:\nchs\respnd95.dat' lrecl=13064;
4 Input
5 TODAYSPG 6847-6847
6 CONSTAT1 11934-11935
7 CONSTAT2 11936-11937
8 CONSTAT3 11938-11939
9 CONSTAT4 11940-11941
10 SEX1MTHD 11945-11946
11 POST_WT 12350-12359;
12 if constat1 = 'ab' then vjvar=1; else vjvar = 2;
13 WGT1000=POST_WT/1000;
14 title 'NSFG cycle 1995';

NOTE: Character values have been converted to numeric values at the places given by: (Line):(Column).
12:15
NOTE: The infile 'd:\nchs\respnd95.dat' is:
File Name=d:\nchs\respnd95.dat,
RECFM=V,LRECL=13064

NOTE: Invalid numeric data, 'ab' , at line 12 column 15.
RULE: ----+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8----+----9----+----0
1 1000000111260837511521 1 1050 12 106921124112411189
101 2
201 19211059110611197


Slide 12
Sample:- Original Log (cont.)


12901 11232521101 05267213103033921811931011103 01030000000321120000392702210611511200403 1344 1316
13001 622501001006034
TODAYSPG=1 CONSTAT1=5 CONSTAT2=88 CONSTAT3=88 CONSTAT4=88 SEX1MTHD=1 POST_WT=2545.7569 vjvar=2 WGT1000=2.5457569 _ERROR_=1
_N_=20
NOTE: 10847 records were read from the infile 'd:\nchs\respnd95.dat'.
The minimum record length was 13064.
The maximum record length was 13064.
NOTE: The data set WORK.ONE has 10847 observations and 9 variables.
NOTE: DATA statement used:
real time 39.88 seconds
cpu time 12.10 seconds

15 proc freq;
16 tables CONSTAT1 vjvar;
17 run;

NOTE: There were 10847 observations read from the data set WORK.ONE.
NOTE: PROCEDURE FREQ used:
real time 0.49 seconds
cpu time 0.04 seconds

Slide 13
Sample:- Cleaned Log

1 options nocenter;
2 Data one;
3 Infile 'd:\nchs\respnd95.dat' lrecl=13064;
4 Input
5 TODAYSPG 6847-6847
6 CONSTAT1 11934-11935
7 CONSTAT2 11936-11937
8 CONSTAT3 11938-11939
9 CONSTAT4 11940-11941
10 SEX1MTHD 11945-11946
11 POST_WT 12350-12359;
12 if constat1 = 'ab' then vjvar=1; else vjvar = 2;
13 WGT1000=POST_WT/1000;
14 title 'NSFG cycle 1995';

NOTE: Character values have been converted to numeric values at the places given by: (Line):(Column).
12:15
NOTE: The infile 'd:\nchs\respnd95.dat' is:
File Name=d:\nchs\respnd95.dat,
RECFM=V,LRECL=13064

NOTE: Invalid numeric data, 'ab' , at line 12 column 15.

Slide 14
Sample:- Cleaned Log (cont.)

NOTE: 10847 records were read from the infile 'd:\nchs\respnd95.dat'.
The minimum record length was 13064.
The maximum record length was 13064.
NOTE: The data set WORK.ONE has 10847 observations and 9 variables.
NOTE: DATA statement used:
real time 39.88 seconds
cpu time 12.10 seconds

15 proc freq;
16 tables CONSTAT1 vjvar;
17 run;

NOTE: There were 10847 observations read from the data set WORK.ONE.
NOTE: PROCEDURE FREQ used:
real time 0.49 seconds
cpu time 0.04 seconds

Slide 15
Forbidden Commands

  • Commands that pose unacceptable disclosure risks OR
  • Disallowed to protect integrity/internal environment of ANDRE
    • Add
    • editor
    • report
    • iml
    • Print first.
    • Pctn
    • nofreq
    • Obs last.
    • Pctsum
    • nocum
    • Firstobs
    • nocol
    • tabulate
    • editor
    • Browse
    • summary
    • list
    • put

Slide 16
Commands Modification

  • Modify users program to enforce restrictions on options allowed with certain SAS procedures to prevent objectionable info appearing in the output

PROC MEANS n mean std;

Slide 17
Output Suppression

  • Wiping out of extreme values from the output of Proc Univariate
  • Suppressing complete output line (Procs Means, corr, Univariate etc) where sample size less than the minimum acceptable value.

Slide 18
Proc Means Suppression

The MEANS Procedure

Variable Label N Mean Std Dev
--------------------------------------------------------------------------------------------
EXPEND_R Current expend/pupil in public schl/1000 5424 5.0830820 1.3958710
*** Values Suppressed ***
 RPUB87 exp. for contr. serv. and supplies 1997$ 5424 23472052.60 18806802.86
RPUB92 exp. for contr. serv. and supplies 1997$ 5424 34800922.98 30481634.59
PRGPRO Coordinated Pregnancy Prevention Program 1708 0.0679157 0.2516749
HIVED HIV/AIDS Education 1708 3.5146370 0.8044378
*** Values Suppressed ***
 PRGPRO87 Coordinated Pregnancy Prevention Program 5424 0.0540192 0.2260764
HIVED87 HIV/AIDS Education 5424 3.4968658 0.8008324
WT_PER15 % Wt females aged 15-19/total 15-19 5424 0.7279681 0.1265796
BK_PER15 % Bk females aged 15-19/total 15-19 5424 0.1409869 0.0932332
HS_PER15 % Hs females aged 15-19/total 15-19 5424 0.0962413 0.1055191
TEENMMC2 Teenmom by cohort (1,2,3r) 1201 1.7119067 0.7715351
C18_2_1S R in C2 (vs 1) at 18-19 endpt (1,2) 1770 1.5248588 0.4995228
TM2_1S18 R tnmm in Coh 2 (vs 1)-age 18 @ ext 358 1.4804469 0.5003168
AGE_12 Date R = 12 in century months 6450 979.5613953 69.3124265
STRTST IA5 Date R started living in current sta 3870 1132.55 753.2066507
BDAYCENM R date of birth 6450 835.5613953 69.3124265
RAVPAY95 real av. an. pay 95 dollars 5424 26933.93 2826.80
PERCAFDC percent of households receiving AFDC 5424 0.0422254 0.0127307
SALARY teacher salaries real 96-97$$$ 5424 35338.66 5729.11
--------------------------------------------------------------------------------------------

Slide 19
Proc Univariate Output Unsuppressed

The SAS System 9
14:09 Sunday, October 24, 1999
Univariate Procedure
Variable=AVHRATET
Moments Quantiles(Def=5)
N 2283 Sum Wgts 2283 100% Max -0.25314 99% -1.62008
Mean -4.66219 Sum -10643.8 75% Q3 -3.56179 95% -2.37588
Std Dev 1.892017 Variance 3.57973 50% Med -4.50491 90% -2.79152
Skewness -2.11919 Kurtosis 6.892929 25% Q1 -5.30374 10% -6.07639
USS 57792.36 CSS 8168.944 0% Min -13.5463 5% -7.19645
CV -40.5821 Std Mean 0.039598 1% -12.7402
T:Mean=0 -117.738 Pr>|T| 0.0001 Range 13.29321
Num ^= 0 2283 Num > 0 0 Q3-Q1 1.741949
M(Sign) -1141.5 Pr>=|M| 0.0001 Mode -13.5463
Sgn Rank -1303593 Pr>=|S| 0.0001

Extremes

Lowest Obs
-13.5463( 1547)
-13.5397( 1836)
-13.4637( 2084)
-13.4413( 1127)
-13.4402( 1088)

Highest Obs
-0.90519( 649)
-0.81756( 1094)
-0.76928( 1739)
-0.5907( 21)
-0.25314( 400)

Slide 20
Proc Univariate Output Suppressed

The SAS System 9
14:09 Sunday, October 24, 1999
Univariate Procedure
Variable=AVHRATET
Moments Quantiles(Def=5)
N 2283 Sum Wgts 2283 100% Max -0.25314 99% -1.62008
Mean -4.66219 Sum -10643.8 75% Q3 -3.56179 95% -2.37588
Std Dev 1.892017 Variance 3.57973 50% Med -4.50491 90% -2.79152
Skewness -2.11919 Kurtosis 6.892929 25% Q1 -5.30374 10% -6.07639
USS 57792.36 CSS 8168.944 0% Min -13.5463 5% -7.19645
CV -40.5821 Std Mean 0.039598 1% -12.7402
T:Mean=0 -117.738 Pr>|T| 0.0001 Range 13.29321
Num ^= 0 2283 Num > 0 0 Q3-Q1 1.741949
M(Sign) -1141.5 Pr>=|M| 0.0001 Mode -13.5463
Sgn Rank -1303593 Pr>=|S| 0.0001

Slide 21
Proc Univariate Output Suppressed (sample size = 1)

Univariate Procedure
Variable=FREQ (sum) freq
Moments Quantiles(Def=5)

Serious Disclosure limitation Violations
Values too low to release
Output of Proc Univariate withheld

Slide 22
Proc Freq Suppression (one way Tables)

  • Suppress at least two consecutive rows to prevent derivation of suppressed values from cumulative totals.
  • Disallow single row output.

Slide 23
1-Way Freq Table suppressed

Cumulative Cumulative
LOGRNTOPAT Frequency Percent Frequency Percent
-----------------------------------------------------------------
0.2277839309 ????? ????? ????? ?????
0.2277839309 ????? ????? ????? ?????
0.2305236586 5 0.08 6429 97.99
0.231111721 5 0.08 6434 98.06
0.232058915 ????? ????? ????? ?????
0.232058915 ????? ????? ????? ?????
0.2436220827 ????? ????? ????? ?????
0.2436220827 ????? ????? ????? ?????
0.2498117984 6 0.09 6456 98.40
0.2504106777 6 0.09 6462 98.49
0.2513144283 18 0.27 6480 98.77
0.2595111955 6 0.09 6486 98.86
0.2670627852 ????? ????? ????? ?????
0.2670627852 ????? ????? ????? ?????
0.2736958305 5 0.08 6500 99.07
0.2814124594 5 0.08 6505 99.15
0.3022808719 6 0.09 6511 99.24
0.3364722366 10 0.15 6521 99.39

Slide 24
1-Way Freq Table suppressed (cont.)

Cumulative Cumulative
LOGRNTOPAT Frequency Percent Frequency Percent
-----------------------------------------------------------------
0.3403258059 ????? ????? ????? ?????
0.3403258059 ????? ????? ????? ?????
0.3715635564 6 0.09 6537 99.63
0.3856624808 ????? ????? ????? ?????
0.3856624808 ????? ????? ????? ?????
0.6931471806 6 0.09 6550 99.83
1.2527629685 ????? ????? ????? ?????
1.2527629685 ????? ????? ????? ?????
1.2527629685 ????? ????? ????? ?????

Slide 25
Proc Freq Suppression (Two way Tables)

  • Rows and columns totals preserved
  • Cells with values less than the acceptable minimum are suppressed
  • Additional suppressions to ensure that no row and no column has single suppression.
  • Logical stitching of horizontal and vertical splits.

Slide 26
Proc Freq: 2-way Tables Suppression

TABLE OF FAMREL BY FAMSIZER
FAMREL FAMSIZER
Frequency
Percent
Row Pct

Col Pct 2 3 4 5 Total
3 94 388 792 533 2206
3 3.97 16.40 33.47 22.53 93.24
3 4.26 17.59 35.90 24.16  
3 98.95 96.28 96.12 94.34  
4 ?????? 9 22 27 104
4 ?????? 0.38 0.93 1.14 4.40
4 ?????? 8.65 21.15 25.96
4 ?????? 2.23 2.67 4.78  
6 ?????? 6 10 5 56
6 ?????? 0.25 0.42 0.21 2.37
6 ?????? 10.71 17.86 8.93  
6 ?????? 1.49 1.21 0.88  
Total 95
4.02		 403
17.03 		824
34.83 		565
23.88 		2366
100.00

Slide 27
Proc Freq: 2-way Tables Suppression (Cont.)

checking frequencies 4
12:01 Thursday, May 6, 1999
TABLE OF FAMREL BY FAMSIZER
FAMREL FAMSIZER
Frequency
Percent
Row Pct

Col Pct 6 7 8 9 Total
3 209 98 19 73 2206
3 8.83 4.14 0.80 3.09 93.24
3 9.47 4.44 0.86 3.31  
3 90.48 83.05 59.38 74.49  
4 13 10 ?????? 12 104
4 0.55 0.42 ?????? 0.51 4.40
4 12.50 9.62 ?????? 11.54  
4 5.63 8.47 ?????? 12.24  
6 9 10 ?????? 13 56
6 0.38 0.42 ?????? 0.55 2.37
6 16.07 17.86 ?????? 23.21  
6 3.90 8.47 ?????? 13.27  
Total 231
9.76 		118
4.99 		32
1.35 		98
4.14 		2366
100.00

Slide 28
Contact Information

For general Questions/Comments
Email : rdca@cdc.gov
Phone: (301) 458-4732

For On-site Info:
Email : Neb9@cdc.gov
Phone: (301) 458-4097

For Remote Access Info:
Email : vgambhir@cdc.gov
Phone: (301) 458-4226