complete the national crime and victimization survey

[ajdamico]

[ajdamico]

code to convert is http://www.bjs.gov/content/pub/pdf/Variance_Guide_Appendix_C_SAS.pdf

[ajdamico]

to match the standard errors in http://www.bjs.gov/content/pub/pdf/Variance_Guide_Appendix_E_SPSS.pdf

our annual bulletin (http://www.bjs.gov/content/pub/pdf/cv14.pdf)

have users focus on replicating percentages or counts rather than rates. With counts and percentages, you only need to use the incident file and should not have an issue with direct estimation of standard errors once you account for the NCVS complex sample design.

the weight variable newwgt gets defined in http://www.bjs.gov/content/pub/pdf/NCVS_Variance_User_Guide%2011.06.14.pdf#page=32

If you just want to replicate the estimates provided in the annual bulletin, you’ll use v4022 to identify crimes that occurred outside of the US and v4529 (1-20) to identify violent crime.

This is what you’ll run to generate a count of violent crime:

missing values all().
recode v4529(1,2,3,4,15,16,18,19=1)   /* Rape & Sexual Assault
            (5,6,7,8,9,10       =2)   /* Robbery
            (11,12,13           =3)   /* Aggravated Assault
            (14,17,20           =4)   /* Simple Assault
            (21,22,23           =5)   /* Personal Larceny
            (31,32,33           =6)   /* Household Burglary
            (40,41              =7)   /* Motor Vehicle Theft
            (54 thru 59         =8)   /* Theft
            into newoff.

recode newoff(1 thru 4=1)             /* Violence
             (5       =2)             /* Personal Larceny
             (6,7,8   =3)             /* Property
             into newcrime.



variable labels
newoff       ''
newcrime     ''.



value labels
newoff     1 'Rape & Sexual Assault'
            2 'Robbery'
            3 'Aggravated Assault'
            4 'Simple Assault'
            5 'Personal Larceny'
            6 'Household Burglary'
            7 'Motor Vehicle Theft'
            8 'Theft'/

newcrime   1 'VIOLENCE'
            2 'PERSONAL LARCENY'
            3 'PROPERTY'/.

execute.
Select if (v4022 ne 1).
Weight by newwgt.
Freq vars newcrime.

That’s the basic code to generate an estimate of violent crime. You should be able to use that to match any violent crime count estimates in the bulletin or in the variance guide appendices. You’ll need to account for the complex sample design to generate accurate standard errors.

Regardless, you should also be able to use direct estimation with the public use files to replicate any of the counts or percents and corresponding standard errors shown in Appendix C. You just won’t be able replicate rates, unless you create the flat file, and you’ll use the variable names from the public use file, rather than creating the new variables that were created for the purpose of adding incidents to the person file.

Again, on the public use incident file, v4022 (1) identifies crimes that occurred outside of the US and v4529 (1-20) identifies violent crime. You can use these variables and others on the public use file and tsl to replicate the Appendix C counts and percentages. The design variables are the same ones described in the report and the Appendix.

We just don’t have anyone in the victimization unit who’s working in SAS so we don’t have SAS code or output that we can share.

[ajdamico]

this one may have a haiku..