4906.0.55.001 - Microdata: Personal Safety, Australia, 2012 Quality Declaration 
ARCHIVED ISSUE Released at 11:30 AM (CANBERRA TIME) 13/05/2014   
   Page tools: Print Print Page Print all pages in this productPrint All

USING THE CURF

This chapter contains the following sections:

FILE STRUCTURE
USING THE FILE LEVELS
MERGING ACROSS LEVELS
WHEN TO USE AGGREGATE DATA ITEMS ON THE PERSON LEVEL
RECORD IDENTIFIERS
COPYING DATA FROM THE PERSON LEVEL TO A LOWER LEVEL
RELIABILITY OF ESTIMATES
NOTES ON DATA ITEMS
CURF FILE NAMES


FILE STRUCTURE

The PSS file is a hierarchical file, made up of seven levels.

At the top level there is a:

    • Person level

Beneath the person level, there are 6 further levels (linked to the person level):
    • Violence prevalence level
    • Violence most recent incident level
    • Violence partner level
    • Emotional abuse by a partner level
    • Abuse before the age of 15 level
    • Stalking level

In using the CURF, it is important to understand how the information was collected as this determines the types of estimates that are able to be produced from the files and which levels on the CURF this information will be found.


USING THE FILE LEVELS

The Person level contains one record for every respondent. The remaining six levels are in a hierarchical relationship to the Person level and can contain multiple records for each respondent, depending on their experiences. As it is possible for a respondent to have more than one record on most levels, there are a number of Index items that are used to uniquely identify each record type. How to use the Index items on each level is outlined further below.


File level
Maximum possible number of records for each respondent on level
Number of records on level
Number of Index items on level

Person level
1
17,050
0
Violence prevalence level
40 (a)
24,475
4
Violence most recent incident level
8
20,551
3
Violence partner level
2
17,087
1
Emotional abuse by a partner level
3
17,259
2
Abuse before the age of 15 level
2
17,759
1
Stalking level
2
17,222
1

(a) excluding totals - see Violence prevalence level.


Person level

This level contains demographic information about the respondent, and where applicable, their partner, along with other information about general feelings of safety and lifestyle indicators, and aggregated data items, including those for sexual harassment. This level contains 1 record for each person. There are 17,050 Person records on this level.

Important notes for using the Person level
    • The person level contains survey weights. Where estimates are derived from the CURF, it is essential that they are calculated using the weight, rather than counting the number of records in each category. For further information on why and how to apply weights, refer to section below on Relative Sampling Error.
    • The person level also contains a set of 60 replicate weights to assist users in the calculation of estimates of the standard error. For further information on why and how to apply replicate weights, refer to section below on Relative Sampling Error.
    • Other information available on the Person level includes aggregated data items. These aggregates provide summary information such as 'Whether experienced any violence since age 15'. These aggregates are summary data items derived from the multiple records held on the different levels. This information has been provided to assist CURF users when deriving simple cross tabulations. For further information refer to section below on When to use Aggregate data items on the person level.
    • The Person level can be used in conjunction with information on the other levels. For example, when looking at a person's experiences of stalking, data on the Stalking level, such as the type of stalking behaviours, can be cross tabulated by information on the Person level, such as sex of the respondent. This is done by merging the person level data onto the Stalking level file. An example of the code to merge data from the Person level to a lower level is provided in Copying data from the person level to a lower level. However please note that the aggregates are not intended to be used in conjunction with data from the same level from which they are derived, for more information about why this is conceptually incorrect, see When to use Aggregate data items on the person level.

Limitations and cautions for producing PSS cross tabulations with the Person level

Data users are advised to consider the following known limitations and cautions when producing any cross tabulations from the PSS CURF Person level:
    • It is not possible to produce reliable estimates from the 2012 PSS for men for Tasmania, the NT and the ACT. This is due to the sample only being designed to produce national level estimates for men. Data users should consider the reliability of any such estimates - see Relative Sampling Error. section below. For further information on Sample design refer to the Sample design and selection section in the Personal Safety Survey, Australia: User Guide, 2012 (cat. no. 4906.0.55.003).
    • It is not possible to produce reliable person level (i.e. men plus women) estimates from the 2012 PSS for Tasmania, the NT and the ACT.
    • Extreme caution should be taken when making inferences about a person's current socio-demographic characteristics and their experience of violence. To assume a person's current socio-demographic characteristics were present at the time of the violence could be highly inaccurate. ABS advises that the best way to assess current socio-demographic characteristics is by a person's recent experiences (i.e. those that occurred in the last 12 months). These are much more likely to match the person's current socio-demographic characteristics rather than assessing their lifetime experiences of violence. For example a person who is now 45 may have last experienced violence as a 21 year old: if we were to assess their current socio-demographics against this, they may have since moved states/areas and may have since developed a disability: these current characteristics may bear no correlation with what would have been their characteristics at the time of the violence). Socio-demographic data items such as state of usual residence, remoteness area, disability status, self perceived health status etc do not remain the same over a person's lifetime. It therefore can not be assumed that these were the person's characteristics at the time of experiencing: violence since the age of 15; emotional abuse by a partner since the age of 15; partner violence since the age of 15; sexual harassment during their lifetime; abuse before the age of 15; or stalking during their lifetime.
A comprehensive list of socio-demographic data items and their response categories and aggregate data items are available from the Downloads tab. For information about socio-demographic data items refer to the following worksheet in the Data Items list - 1 Person. For more information about what aggregate data items are available, refer to the following worksheet in the Data Items list - 1a Aggregates.

Using the lower levels

Beneath the person level, there are 6 further levels (linked to the person level):
    • Violence prevalence level
    • Violence most recent incident level
    • Violence partner level
    • Emotional abuse by a partner level
    • Abuse before the age of 15 level
    • Stalking level
    Each sub level contains different information about all the different types of experiences measured in the PSS. When using these lower levels of the PSS CURF, there are a numbers of factors to consider:
      • In order to identify the different types of experiences on the sub levels for each person, Index items must be specified when extracting information off each level. Where a person has not experienced the type of information contained on a sub-level, they will have 1 record on the level with 0 Not applicable for all Index and data items. Where a person has experienced one or more of the type/s in a level, they will have one or more records on the level for each different type they have experienced and the index items must be used to identify the type of experience.
      • 'Totals' are generally NOT included on each level. Conceptually it is invalid to add together data about the characteristics for the different types, as actions a person may take could differ between incidents depending on the type of violence experienced. In addition, where a person has had multiple experiences, to add data for the different types together, it would result in double counting persons who had experienced more than one type.
      • Data about a person's experiences must be analysed/presented separately for each of the different types on each level.
      On overview of how to use each sub level of the CURF is provided below. More detailed information for each level is provided in Appendix 1 - Detailed descriptions for using each level of the file.

      Data users are strongly encouraged to refer to the Interpretation of Results, Survey Definitions and Data Collection chapters of the Personal Safety Survey, Australia: User Guide, 2012 (cat. no. 4906.0.55.003) for further information to assist them with using information from the PSS CURF.

      Violence prevalence level

      This level contains information about a person's experience of violence since the age of 15 for each of the 40 different types of violence and when their most recent incident of each of these types occurred. The 40 different types of violence include: the 4 different types of violence (ie sexual assault, sexual threat, physical assault & physical threat), by sex of perpetrator (ie by male and female perpetrators), by 5 key perpetrator types (stranger; boy/girlfriend or date; current partner; previous partner; other known person).

      To assist CURF users 'Totals' have been included for each of the Index items on the Prevalence level. Each 'Total' only counts each person once. Where are person has experienced violence they will have a minimum of 16 records on this level.

      Conceptually it is not recommended that information from the Violence prevalence level be merged with data on the other levels (other than the Person level). More information is provided below in Merging across levels.

      Other information about a person's experience of violence is also contained on the Person level as summary aggregated data items. These aggregates provide summary information from each of the different levels (eg 'Whether ever experienced violence' - ie any of the 40 types of violence), however care should be taken when using these items in conjunction with any of the other levels. More information is provided below in When to use Aggregate data items on the person level.

      A list of the index and data items and their response categories on this level are available from the Downloads tab. For further information refer to the following worksheet in the Data Items list - 2 Violence-Prevalence.

      Violence most recent incident level

      The Violence most recent incident level contains information about the characteristics of a person's most recent incident of each of 8 types of violence (ie Sexual assault by a male; Sexual assault by a female; Sexual threat by a male; Sexual threat by a female; Physical assault by a male; Physical assault by a female; Physical threat by a male; and Physical threat by a female). Data about the characteristics of a person's most recent incident must be analysed separately for each of the 8 different types of violence.

      This level does not include information about the characteristics of a persons most recent incident where the incident occurred more than 20 years ago.

      Information on this level is able to be used to analyse the 8 different types of violence to assess:
        • Whether there are differences between what happens when different types of violence are experienced; and
        • Whether there are differences between what happens when a woman experiences violence, and when a man experiences violence.
        Conceptually it is not recommended that information from the Violence most recent incident level be merged with data on the other levels (other than the Person level). More information is provided below in Merging across levels.

        Other information about a person's experience of violence is also contained on the Person level as summary aggregated data items. These aggregates provide summary information from each of the different levels (eg 'Experience of incidents of sexual assault' and whether they had experienced 'Multiple incidents of sexual assault all by same person sexual assault' or 'Multiple incidents of sexual assault by different persons), however care should be taken when using these items in conjunction with the Violence most recent incident level and any levels. More information is provided below in When to use Aggregate data items on the person level.

        A comprehensive list of the index items, the data items and their response categories are available from the Downloads tab. For further information refer to the following worksheet in the Data Items list - 4 Violence - MRI.

        Violence partner level

        The Violence partner level contains information about the characteristics of a person's experience of violence by current partner and/or their most recently violent previous partner. Data about a person's experience of partner violence must be analysed separately for current and previous partner.

        This provides a powerful dataset with comprehensive information for what happens when a person experiences:
          • violence by a current partner; and
          • violence by a previous partner.

        Conceptually it is generally not recommended that information from the Violence partner level be merged with data on the other levels (other than the Person level). More information is provided below in Merging across levels.

        Other information about a person's experience of violence is also contained on the Person level as summary aggregated data items. These aggregates provide summary information from each of the different levels (eg 'Whether experienced current and/or previous partner violence since age 15'), however care should be taken when using these items in conjunction with the Violence partner level or any other levels. More information is provided below in When to use Aggregate data items on the person level.

        A comprehensive list of the index items, the data items and their response categories are available from the Downloads tab. For further information refer to the following worksheet in the Data Items list - 5 Violence-Partner.

        Emotional abuse by a partner level

        The Emotional abuse level is used to assess a person's experience of emotional abuse by current partner and/or male previous partner and/or female previous partner. Data about the characteristics of a person's experience of emotional abuse by a previous partner must be analysed separately for sex of partner (ie male/female previous partner).

        Information on this level is able to be used to analyse for differences in what happens when a person experiences emotional abuse by different partner types.

        Conceptually it is generally not recommended that information from the Emotional abuse by a partner level be merged with data on the other levels (other than the Person level). More information is provided below in Merging across levels.

        Other information about a person's experience of emotional abuse and violence by a partner is also contained on the Person level as summary aggregated data items. These aggregates provide summary information from each of the different levels (eg 'Whether experienced emotional abuse by a current and/or previous partner since age 15'), however care should be taken when using these aggregate data items in conjunction with the Emotional abuse by a partner level or any other levels. More information is provided below in When to use Aggregate data items on the person level.

        A comprehensive list of the index items, the data items and their response categories are available from the Downloads tab. For further information refer to the following worksheet in the Data Items list - 6 Emotional Abuse.

        Abuse before the age of 15 level

        The Abuse before the age of 15 level is used to assess a person's experience of physical and/or sexual abuse by an adult before the age of 15. Data about the characteristics of a person's experience of abuse before the age of 15 must be analysed separately for each type of abuse (ie physical/sexual).

        Where a person has refused to answer the questions regarding their experience of physical and/or sexual abuse before the age of 15 they will have one record on this level for each different type/s of abuse they refused to answer with 3 Physical abuse before age 15 - refused and/or 4 Sexual abuse before age 15 - refused for the index items and 0 Not applicable for all the data items.

        Other related information about a person's experience of abuse before the age of 15 is also contained on the Person level as summary aggregated data items. These aggregates provide summary information (eg 'Whether experienced abuse before age 15' - ie either sexual or physical abuse), and should never be used in conjunction with characteristics from the Abuse before the age of 15 level. More information is provided below in When to use Aggregate data items on the person level.
          A comprehensive list of the index items, the data items and their response categories are available from the Downloads tab. For further information refer to the following worksheet in the Data Items list - 7 Abuse.

          Stalking level

          The Stalking level is used to assess a person's experience of stalking by a male and stalking by a female. Data about the characteristics of a person's experience of stalking must be analysed separately for each sex of perpetrator (ie stalking by a male/stalking by a female).

          Information on this level is able to be used to analyse for differences in what happens when a person experiences stalking by a male or by a female stalker.

          Conceptually it is not recommended that information from the Stalking level be merged with data on the other levels (other than the Person level). More information is provided below in Merging across levels.

          Other information about a person's experience of stalking is also contained on the Person level as summary aggregated data items. These aggregates provide summary information from each of the different levels (eg 'Whether ever experienced stalking' - ie by a male or by a female), however care should be taken when using these aggregate data items in conjunction with the Stalking level or any other levels. More information is provided below in When to use Aggregate data items on the person level.

          A comprehensive list of the index items, the data items and their response categories are available from the Downloads tab. For further information refer to the following worksheet in the Data Items list - 9 Stalking.


          MERGING ACROSS LEVELS

          Besides the information contained on the Person level, each of the different levels of the file are mutually exclusive and in most instances it would be conceptually inaccurate to attempt to merge characteristics, sum or cross tabulate data items across the levels.

          A number of Aggregate data items have been created for use on the Person level which are designed to assist CURF users with their analysis of the co-occurence of different types of violence, abuse, stalking or sexual harassment. These aggregate items should only be used within the Person level and not in conjunction with characteristics data or cross tabulated with data from the lower levels. More information is provided below in When to use Aggregate data items on the person level.

          Where it is conceptually possible to merge data across levels, this has been outlined below. This section also provides a brief description about why it would be conceptually invalid to merge data across certain levels.

          Combining data across the various violence levels

          The three levels that contain information regarding violence are mutually exclusive, that is, data cannot be merged or summed across these levels. For example it would be conceptually inaccurate to combine information about the characteristics of a person's most recent incident on the Violence - Most recent incident level with data about the characteristics of previous partner violence on the Violence - Partner level. Information on the Violence - Most recent incident level, such as Whether injured in most recent incident of physical assault by a male, only relates to a person's most recent incident: even if this was by a male previous partner, this is not whether they were 'ever injured' by a male previous partner (they may not have been injured in the most recent incident but were in an earlier incident). Additionally, as a person may have experienced violence by more than one previous partner, information on the Violence - Most recent incident level may contain information about incidents by different previous partners: hence is not clear whether the information on the Violence - Most recent incident level about male previous partner violence is the same previous partner as that on the Violence - Partner level. Therefore it is not possible to use this information to make statements about whether persons are injured due to partner violence.

          However, certain aggregated data items may assist CURF users with their analysis on the co-occurence of different types of violence. These aggregates provide information on the prevalence of the different types of violence and co-occurrence of different types of violence but care should be taken when using these in conjunction with any violence level. More information is provided below in When to use Aggregate data items on the person level.

          Combining data across the partner violence and emotional abuse by a partner levels

          Current Partner

          Data on these levels is able to be assessed for a person's experience of current partner violence and emotional abuse. This is because a person can only have one current partner: therefore we can assume that the information on the Partner level about current partner violence refers to the same person on the emotional abuse level about current partner emotional abuse.

          Other related information about a person's experience of current partner violence and current partner emotional abuse is also contained on the Person level as summary aggregated data items. These aggregates provide information on the co-occurrence of current partner violence and emotional abuse but should never be used in conjunction with characteristics from the emotional abuse level or partner level. More information is provided below in When to use Aggregate data items on the person level.

          Previous Partner

          Data on the Partner and Emotional abuse levels about previous partner violence and previous partner emotional abuse are mutually exclusive and data cannot be merged or summed across these levels. The previous partner who most recently emotionally abused the respondent is not necessarily the same previous partner who was most recently violent. Information on previous partner emotional abuse cannot be directly linked to previous partner violence. Thus, there are not any aggregates provided for this information on the person level.

          Combining data across the partner violence and stalking levels

          Information collected about stalking in the PSS only relates to the characteristics of a person's most recent episode of stalking. PSS does not have information about the prevalence of stalking by different perpetrator types such as whether ever stalked by a current partner; or whether ever stalked by a previous partner. As information regarding the prevalence of stalking by a partner is not available, combining information from the Violence levels and the Stalking level would underestimate the number of persons who have experienced both violence and stalking by a partner.

          While it is possible to assess for the co-occurrence of both violence and stalking by a current partner for some respondents (whose most recent episode of stalking was by a current partner), this is not a true reflection of all those who have experienced violence and stalking by a current partner and will underestimate those who have experienced both.

          Combining data for previous partner across the partner violence, emotional abuse by a partner and stalking levels

          Similarly while it is possible to assess for the co-occurrence of both violence and stalking by a previous partner for some respondents (whose most recent episode of stalking was by a previous partner), this is not a true reflection of all those who have experienced violence and stalking by a previous partner and will underestimate those who have experienced both. In addition, the previous partner who most recently stalked the respondent is not necessarily the same previous partner who was most recently violent, nor the most recently emotionally abusive previous partner. Information collected about previous partner stalking cannot be directly linked to previous partner violence, or previous partner emotional abuse data.

          Other related information about a person's experience of stalking is also contained on the Person level as summary aggregated data items. These aggregates provide summary information (eg 'Whether ever experienced stalking'), and should never be used in conjunction with characteristics from the stalking level. More information is provided below in When to use Aggregate data items on the person level.


          WHEN TO USE AGGREGATE DATA ITEMS ON THE PERSON LEVEL

          A number of high level aggregate data items have been included on the Person level. The purpose of these aggregate items is to aid CURF users when producing simple cross tabulations about a person's experience of violence/abuse/stalking/sexual harassment by socio-demographic characteristics or to look at the co-occurrence of experiences without having to merge information from each level of the file. An outline of the items and their use is provided below, Aggregate data items are listed in the Data Items list - 1a Aggregates, available from the Downloads tab.

          Abuse before the age of 15 aggregates
          A note of caution: cross tabulating these abuse aggregates with information on the Abuse level would be invalid and would result in double counting (where a person has experienced both sexual and physical abuse). Users wanting to undertake more detailed analysis about the characteristics of abuse before the age of 15 need to use the data on the Abuse level.

          Sexual harassment aggregates

          The additional information collected in the survey regarding sexual harassment has not been included on the CURF. For those wanting to undertake further analysis of this sexual harassment data, additional information may be made available by request, on a fee for service basis, through the ABS Information Consultancy.

          Stalking aggregates

          A note of caution: cross tabulating these stalking aggregates with information on the Stalking level would be invalid and would result in double counting (where a person has experienced stalking by a male and by female). Users wanting to undertake more detailed analysis about the characteristics of Stalking need to use the data on Stalking level.

          Violence aggregates

          There are numerous aggregates that have been included about a person's experience of violence. These are based on information on the Violence Prevalence level and include items such as 'Whether experienced any violence since age 15' (that is whether they have experienced any of the 40 different types of violence since the age of 15)

          For the PSS aggregate data items, where a person has experienced more than one type of violence, they are counted separately for each type of violence they experience and only counted once in any aggregated totals. These aggregates can readily be cross tabulated with the other items on the person level to produce high level summary tables.

          A note of caution: cross tabulating these violence aggregates with information on the Violence - Most recent incident level or Violence Partner level would be invalid and would result in double counting (where a person has experienced violence more than once). Users wanting to undertake more detailed analysis about the characteristics of violence need to use the data on each relevant level for Violence Most recent incident level or Violence Partner level.

          Multiple experiences of violence & co-occurrence aggregates

          There are a number of suites of aggregates that have been included about whether a person has had multiple experiences of violence of various types. Such as 'Whether experienced violence more than once' (ie whether a person has experienced one incident of violence or more than one incident of violence)' and 'Experience of incidents of sexual assault' (ie whether a person has experienced multiple incidents by the same person or multiple incidents by different people). These aggregates have been created using complex derives.

          A suite of aggregates have also been created to look at a persons experience of both violence and/or emotional abuse by their current partner such as 'Experienced current partner violence or current partner emotional abuse since age 15'. Similar aggregates have also been created for only those who have a current partner. This is based on information contained on the Violence - Prevalence level and the Emotional Abuse by a partner level.

          A note of caution: cross tabulating these multiple experiences of violence aggregates with information on the Violence - Most recent incident level or Violence Partner or Emotional abuse by a partner level would be invalid and would result in double counting (where a person has experienced violence more than once). Users wanting to undertake more detailed analysis about the characteristics of violence need to use the data on each relevant level for Violence Most recent incident level or Violence Partner level or Emotional abuse by a partner level.


          RECORD IDENTIFIERS

          Identifiers can be used on records at each level of the file to merge information from the person level to the lower levels. Each person has a unique random identifier – ABSHHID, which appears on all levels of the file.


          COPYING DATA FROM THE PERSON LEVEL TO A LOWER LEVEL

          There may be instances when a data item is not contained on the level of the file required by users. For example, information about the characteristics of a person's eight most recent experiences of violence is contained on the Violence most recent incident level, while that person's geographic and demographic information is contained on the Person level. To apply Person level characteristics to the Violence most recent incident level, the Person level characteristics (on the Person level file PSS12PER) must be merged on to the Violence most recent incident level file (PSS12MRI) using the person identifier (ABSHHID). To do this, the following SAS code (or equivalent) can be used:
            PROC SORT DATA = PSS12PER;

            BY ABSHHID;

            PROC SORT DATA = PSS12MRI;

            BY ABSHHID;

            DATA MERGEFILE;

            MERGE PSS12MRI (IN = A)

            PSS12PER (KEEP = ABSHHID AGECF SEX PSSFINWT IN = B);

            BY ABSHHID;

            RUN;

          The KEEP statement includes all person data items specified to be merged onto the Violence most recent incident level file. Person characteristics can be merged on to the Violence most recent incident level to identify the characteristics of the people who reported one or more violent experiences since the age of 15, and these characteristics are applied to each 'type of violence' record on the Violence most recent incident level. In the above example, age (item AGECF), sex (item SEX) and the person weight (item PSSFINWT) are merged onto the Violence most recent incident level.


          RELIABILITY OF ESTIMATES

          Use of weights

          As the survey was conducted on a sample of private dwellings in Australia, it is important to take account of the method of sample selection when deriving estimates from the CURF. This is particularly important for the 2012 PSS as:
            • there was a high level of non-response due to the voluntary nature of the survey;
            • only one person over the age of 18 was selected in each household; and
            • a larger proportion of females living in smaller states were selected compared to the actual proportions in the population (driven by the requirement for key estimates of women at the State and Territory level).

          Every person has been assigned a 'weight'. The identifier is PSSFINWT, which can be found on the Person level file. The weight indicates how many people in the in-scope population are represented by the selected person. Information for sampled persons must be multiplied by the weights to produce estimates for the whole population. If each person's weight were to be ignored when analysing the data to draw inferences about the population, then no account would be taken of each person's differing chance of selection or of different response rates across population groups, with the result that the estimates produced could be seriously biased. The application of weights ensures that estimates will conform to an independently estimated distribution of the population by certain categories including state, age and sex rather than to the distributions within the sample itself.

          It should be noted that as a result of some of the changes made to protect confidentiality on the CURF following the calculation of weights, estimates of benchmarked items produced from the CURF may not equal the benchmarked values. For further information refer to the Reconciliation of the CURF data document in this product.

          For further information on the sample design and weighting process, refer to Sample Design and Selection and Weighting and estimation procedures in Personal Safety Survey, Australia: User Guide, 2012 (cat. no. 4906.0.55.003).

          Relative sampling error

          Two types of error are possible in an estimate based on a sample survey: non sampling error and sampling error. It is important that the sampling error is taken into account in any data analysis. For further information on non-sampling and sampling error refer to Data Quality chapter of Personal Safety Survey, Australia: User Guide, 2012 (cat. no. 4906.0.55.003). It is also recommended that data users conduct significance testing when analysing for any differences between estimates. For further information refer to Significance testing on differences between survey estimates in the Data Quality chapter of Personal Safety Survey, Australia: User Guide, 2012 (cat. no. 4906.0.55.003).

          Associated with each person record on the CURF is a set of 60 'replicate weights' (WHM0201 to WHM0260) in addition to the 'main weight' (PSSFINWT). These replicate weights can be used to derive estimates of standard error.

          The basic idea behind the replication approach is to select subsamples repeatedly (60 times) from the whole sample and re-weight each sub-sample to align with the benchmarks. The resultant weights are referred to as replicated weights. For each of these subsamples the statistic of interest are calculated using the associated replicate weight. The variance of the full sample statistic is then estimated using the variability among the replicate statistics calculated from the subsamples. As well as enabling variances of estimates to be calculated relatively simply, replicate weights also enable unit record analyses such as chi–square tests and logistic regression to be conducted which take into account the complex sample design. There are various ways of creating replicate subsamples from the full sample. The replicate weights produced for the 2012 PSS have been created using a group jackknife method of replication.

          The formulae for calculating the SE and RSE of an estimate using the group jackknife replicate weights are:

            SE(y) = square root of (59/60) X [(yg1 - y) squared + (yg2 - y)squared + ... continue until (yg60 - y) squared] . Definitions of "yg" and "y" are provided textually below.
            where

            g = 1,..,60 (the no. of replicate groups)

            y(g) = weighted estimate, having applied the weights for replicate group g

            y = weighted estimate from the full sample weight.

            RSE(y) = SE(y)/y * 100%.


          To enable CURF users to check that they are using the replicate weights correctly, an example table has been included on the RADL: Example Tables PSS12 CURF.xls . This document includes an example table (for each level of the file) of: the weighted estimates; the sample counts; and the relative standards errors calculated from the CURF for at least one item on each level. This document will allow CURF users to check that they are extracting the correct data for any particular level from the CURF and are calculating the associated RSEs correctly.


          NOTES ON DATA ITEMS

          The data items available on the CURF, and the categories within the data items, can differ in detail from those output from the survey (available through an ABS Information Consultancy). The data items list also shows the differences between what is available from the survey and what has been released on the CURF. Many of the differences result from the need to ensure confidentiality of information released on the CURF.

          A complete list of the data items available on each record level for the CURF, including relevant population and classification details, is available from the Downloads tab.

          Many of the data items included on the CURF are self-explanatory. The Glossary in Personal Safety Survey, Australia: User Guide, 2012 (cat. no. 4906.0.55.003) provides links to terms and definitions for most of the survey's data items. However, some items require further explanation, which has been provided below.

          Not known and Refusal responses

          The survey allowed respondents the option to 'refuse' to answer particularly sensitive questions. It also allowed for a 'Not known' response where they were unable to provide an answer (eg if they did not know a response to questions about their partner's income, education, etc).

          Where this was allowable, data items will have a specific response category shown as either: 'Don't know' 'Refusal' 'Don't know/refusal' or 'Not stated'. Refer to the values used for each of the Response categories in the Data Items list.

          When using data items that contain 'Not known/refusal' responses, this should be considered in analysis and noted in any outputs. One option is to list these responses as a separate output category and note as such.

          Not applicable responses

          Where an item is not relevant to a particular respondent or subpopulation they will be included in a data item as a 'Not applicable' response.

          Where not applicable responses are present, this means that the data item is only applicable to a sub population or is not relevant to all persons in the subpopulation and should be described appropriately in any outputs. For example, data items relating to current partner violence should be described as a characteristic of those 'who experienced current partner violence', rather than a characteristic of all persons. Data about type of injuries received should be described as characteristics of 'those injured in their most recent incident of physical assault by a male', rather than characteristics of all persons.

          Multiple response categories

          The PSS contains a number of data items where a respondent was able to select more than one response category eg 'All reasons for not contacting the police'. Where data items allowed for more than one response to be given they are noted on the Data Items list as a 'Multi-response item' directly beneath the data item name.

          When using multiple response data items the total number of responses for outputs will be greater than the total number of respondents. Where data items are multiple response, it also means that the components are not able to be aggregated to form a total. An example of such an item from the Violence most recent incident level is 'MRI Physical assault behaviours experienced'. As it is possible that a person has experienced more than one type of physical assault behaviour, they will be counted separately for each type of behaviour experienced. If CURF users were to aggregate the total number of behaviours, this would double count those persons who experienced more than one type of physical assault behaviour.


          CURF FILE NAMES

          The 2012 PSS Expanded CURF can be accessed on the RADL or ABSDL and is available in SAS, SPSS and STATA formats and comprise the following files:
            • Data files
            • Frequency files
            • Information files

          Data files

          SAS files:

          These files contain the data for the CURF in SAS for Windows format.
            • PSS12PER.sas7bdat contains the Person level data
            • PSS12PRV.sas7bdat contains the Violence prevalence level data
            • PSS12MRI.sas7bdat contains the Violence most recent incident level data
            • \PSS12PAR.sas7bdat contains the Violence partner level data
            • PSS12EMAB.sas7bdat contains the Emotional abuse by a partner level data
            • PSS12ABU.sas7bdat contains the Abuse before the age of 15 level data
            • PSS12STK.sas7bdat contains the Stalking level data

          SPSS files:

          These files contain the data for the CURF in SPSS for Windows format.
            • PSS12PER.sav contains the Person level data
            • PSS12PRV.sav contains the Violence prevalence level data
            • PSS12MRI.sav contains the Violence most recent incident level data
            • PSS12PAR.sav contains the Violence partner level data
            • PSS12EMAB.sav contains the Emotional abuse by a partner level data
            • PSS12ABU.sav contains the Abuse before the age of 15 level data
            • PSS12STK.sav contains the Stalking level data

          STATA files:

          These files contain the data for the CURF in STATA format.
            • PSS12PER.dta contains the Person level data
            • PSS12PRV.dta contains the Violence prevalence level data
            • PSS12MRI.dta contains the Violence most recent incident level data
            • PSS12PAR.dta contains the Violence partner level data
            • PSS12EMAB.dta contains the Emotional abuse by a partner level data
            • PSS12ABU.dta contains the Abuse before the age of 15 level data
            • PSS12STK.dta contains the Stalking level data
          Frequency files:

          The following files contain documentation about data item code values and category labels at each level, with weighted and unweighted frequencies for each value.
            • FREQUENCIES_PSS12PER.xls contains the Person level data
            • FREQUENCIES_PSS12PRV.xls contains the Violence prevalence level data
            • FREQUENCIES_PSS12MRI.xls contains the Violence most recent incident level data
            • FREQUENCIES_PSS12PAR.xls contains the Violence partner level data
            • FREQUENCIES_PSS12EMAB.xls contains the Emotional abuse by a partner level data
            • FREQUENCIES_PSS12ABU.xls contains the Abuse before the age of 15 level data
            • FREQUENCIES_PSS12STK.xls contains the Stalking level data

          Information files:

          There are a number of information files associated with the 2012 PSS Expanded CURF:
            • PSS12 Final DIL MURF_CURF.xls is the PSS 2012 data item list and contains all the data items, including details of categories and code values, that are available on the Expanded CURF. This is available from the Downloads tab.
            • FORMATS.sas7bcat is a SAS library containing formats.
            • Example Tables PSS12 CURF.xls This document includes an example table for each level of the file of: the weighted estimates; the sample counts; and the relative standards errors calculated from the CURF for at least one item on each level. This document will allow CURF users to check that they are extracting the correct data for any particular level from the CURF and are calculating the associated RSEs correctly.