[Voiceover]
National Data Archive on Child Abuse and Neglect. 

[Erin McCauley]
So Hi everyone we are so pleased that you are here to join us. This is I believe the third annual NDACAN Summer Training Series where we try to give workshops or presentations about NDACAN data or statistical analysis. We do turn all of Summer Training Series into a webinar series when we're done, and those will be posted on our website. So if you saw something that you thought was interesting you want to come back to or you think your colleague may like, you know, we recommend checking out the series once it's on-line. So as I said this is the NDACAN Summer Training Series. It's run by the National Data Archive on Child Abuse and Neglect which is affiliated with both Cornell University and Duke University. The theme of this summer is Data Strategies for the Study of Child Welfare. Last week we had an introduction to NDACAN and this week we're jumping into our survey-based and cross-sectional datasets. So here's an overview of the whole summer. So we are on July 14, survey-based data. Just as a little preview for where the workshop workshops are going for the rest of the summer we have next week we're going to have a little overview of the administrative data and then go over the process of linking those data. The following week we'll be talking about the VCIS data which is our one of our newer datasets that allows us to extend the panel of youth involved in child welfare back further in time. So that'll be exciting. Then we have an analysis workshop on how to do multilevel modeling so we'll talk about kind of the theory and also go through an example using our data. And then we'll round out the summer with a latent class analysis workshop. Similarly we'll be going over the theory and then going through an example using our data. I'm now going to pass it over to Holly who is our lovely presenter for today. She is the data analyst for the survey-based data products. If you've been to our office hours or sent in questions about these data you will probably have heard from her before. But I realize I forgot to introduce myself. So I am Erin McCauley the graduate research associate with NDACAN. I do a lot of the public-facing workshops and sessions so if you've seen us at perhaps a conference workshop you will have met me. So I'm going to pass it over to Holly and we will get started. 

[Holly Larrabee]
So today’s Session Agenda. First I’d like to say welcome to this session about survey-based data at NDACAN. My name as Erin pointed out is Holly Larrabee and I am a Statistician here at the Archive. I’ve been working at the Archive for the past 17 ˝ years and am the content expert for many of our survey-based datasets. The content of this session will serve as a VERY basic introduction to the National Survey on Child and Adolescent Well-being 1 and 2, the Longitudinal Studies on Child Abuse and Neglect, and the National Child Welfare Workforce Initiative Comprehensive Organizational Health Assessment 2. I am only able to scratch the surface of each one of these datasets in order to stay within the time allotted for this session. The reason for focusing on these datasets out of our entire holdings of survey-based data is because the first two, NSCAW and LONGSCAN, represent our most-popular survey-based datasets and the last one is a forthcoming dataset that will be released soon, so it’s sort of a sneak peek. 
And our first dataset is going to be the National Survey on Child and Adolescent Well-being. The NSCAW 1 refers to the older data collection and NSCAW 2 will refer to the most recent archived data collection effort. 
What is NSCAW? Both of the NSCAW datasets are a nationally representative, longitudinal study of children and families coming into contact with the child welfare system. 
The studies intended to answer a range of fundamental questions about the outcomes for abused and neglected children and their involvement in the child welfare system.
NSCAW 1 was the first national study of child welfare to collect data from children and families, and the first to relate child and family well-being to family characteristics, experience with the child welfare system, community environment, and other factors. 
The Investigators and Sponsors of NSCAW 1 and NSCAW 2
NSCAW was sponsored by the U.S. Department of Health and Human Services, Administration for Children and Families (ACF) and conducted by RTI International as the lead organization in collaboration with the other organizations listed on this slide. 
For NSCAW 1 we had collaborations with the University of North Carolina at Chapel Hill, ICF Caliber, University of California at Berkeley, and Child and Adolescent Services Research Center at San Diego Children’s Hospital collaborated on both NSCAW 1 and NSCAW 2.
And then for NSCAW 2 we also had Walter R. McDonald and Associates, Tufts-New England Medical Center, and Children and Family Research Center, University of Illinois 
NSCAW Versions. There are two versions of each of the NSCAW datasets: there’s a Restricted Release and a General Release.
This presentation will only focus on the Restricted Release version
A brief reason for this is because the general use version of the dataset is heavily redacted and missing two of the variables necessary to account for the complex sample design. I’ll touch on this stuff a little bit later. 
NSCAW 1 Sampling Design Overview. We cannot talk about NSCAW without diving into the sample design which can be a little intimidating if this is your first time kind of working with or hearing about weighted data. The NSCAW study employed a two-stage stratified sampling design. The first stage divided the U.S. into nine sampling strata. Each of the strata correspond to the eight largest participating states. These states are identified by name in the NSCAW 1 dataset but are masked in NSCAW 2. The ninth stratum consists of the remaining 38 states and the District of Columbia and are identified in the dataset as “Remainder” stratum.
Within each of the nine strata, primary sampling units were formed and selected.
A PSU is defined as geographic areas that encompass the population served by a single child protective service agency.
There are nuances about the PSUs that are not covered in this presentation that you will need to become aware of prior to analyzing the data. And that really goes for many of the topics that we’ll cover is that these are covered in more detail in the actual documentation for the dataset. NSCAW 1 Overall Sample Description. The NSCAW 1 sample consists of 6,228 children ages birth to 14 years old, who had contact with the child welfare system within a 15-month period beginning in October, 1999. There are two cohorts in NSCAW one, the Child Protective Service or CPS cohort and the Long Term Foster Care or LTFC cohort. In the Child Protective Services sample, there are 5,501 children, that’s an unweighted count, who were the subject of a child abuse or neglect investigation conducted by CPS during the period between October 1999 and December 2000.
In the Long Term Foster Care sample, there are 727 children again an unweighted count, who had been in out-of-home care for approximately one year at the time of sampling and whose placement had been preceded by a CPS investigation
Please Note that children from states requiring agency first contact were excluded from the study. Agency first contact means that the state required the agency to reach out to eligible study participants first before study staff could contact them. And there’s an entire section again in the documentation that describes why they decided to not I guess recruit children from those states.
The NSCAW 1 CPS Sampling Specifics. The following groups were oversampled to obtain sufficient observations for sub-group analyses. Infants were oversampled to ensure there would be enough cases through to permanency planning. Sexual abuse cases were oversampled to ensure sufficient statistical power to analyze this type of abuse alone. Cases receiving ongoing services after investigation were oversampled to ensure adequate power to understand the process of services. The age of children at investigation was capped at 14 years of age to increase the likelihood that the youth could be located throughout the course of the study. NSCAW 1 Long Term Foster Care Sampling Specifics. 
Children were placed into out-of-home care approximately 12 months before the sample selection period. This means that the child may have been in out-of-home care at the beginning and end of the sampling time period, but may not have been in out-of-home care continuously. It turns out that the sampling selection was later extended due to small number of children qualifying, this resulted in children having spent between 8 and 20 months in out-of-home care at the time of sampling. Placement into out-of-home care was preceded by an investigation for assessment of child abuse or neglect or by a period of in-home services. NSCAW Interview Respondents. As mentioned at the start of this presentation, NSCAW sought to relate child and family well-being to family characteristics, experience with the child welfare system, community environment, and other factors. The way in which they went about that was to collect data from multiple respondents including: the Child, which included Emancipated Youth and Young Adults, the Current Caregiver which included both permanent and non-permanent caregivers, the Caseworker which included the Investigative caseworker and services caseworker, the child’s teacher, and the local agency director or their designee. NSCAW 1-CPS Sample: Data collection Timing. This table shows the dates of data collection and the span of time between waves of data collection. The top row of information contains the data collection start and end dates for each of the waves. The bottom row shows the number of months after the close of the CPS investigation, that qualified them for the NSCAW study, to the specific data collection wave. There were 2 to 6 months from close of investigation to wave 1, 12 months to wave 2, 18 months to wave 3, 36 months to wave 4, and 59 to 96 months to wave 5. 
The study spanned between the close of the CPS investigation that qualified them for the study and wave 5 was about 5 to 8 years, depending upon the timing of the wave 5 interview, which I will talk about more in the next slide. NSCAW 1 CPS Sample: data collection Notes. There was no Wave 2 child or teacher interview. Local Agency Director interview was only conducted at Wave 1. Wave 5 was atypical, in that, it was fielded by age cohort rather than the time interval since the close of investigation that qualified them for the study. So the Infant Cohort was fielded first in September 2005 through February 2006. Children 12 to 48 months old at time of sampling were interviewed for wave 5 in February 2006 to November 2006. Young Adults who had turned 18 years old by April 30, 2006 were next in line for the wave 5 data collection with the remaining cases being fielded March 2007 to December 2007. NSCAW 1 Long Term Foster Care Sample: Data Collection Timing. The table on this slide shows the number of months between being placed in out-of-home care and the waves of data collection. There were approximately 12 months between being placed in out-of-home care and wave 1, 24 months to wave 2, 30 months to wave 3, and 48 months to wave 4. 
NSCAW 1 Long Term Foster Care Sample: Data Collection Notes.
As was the case with the Child Protective Services sample, the Local Agency Director Interview was only conducted at Wave 1 for the Long Term Foster Care cohort. There was no wave 5 interview. 
An interesting note: The unweighted sample size for the long term foster care sample is 727 children. Researchers interested in studying children in long-term foster care meeting similar inclusion criteria for having been in foster care for a period of at least 12 months can construct a larger unweighted sample using children from the CPS cohort. So if you’re interested in examining foster care and are thinking that the long-term foster care sample would be good to use. It may actually be more beneficial to take a peek at the CPS sample because you may end up with a larger sample size overall.
NSCAW 1 and 2: Constructs Measured. There are too many constructs for me to list or discuss in this presentation. A list of constructs measured in NSCAW 1 and NSCAW 2 is available on the NSCAW 1 and 2 Restricted Release datasets pages of the NDACAN website. For your convenience, a link to the document has also been posted in the chat window of this presentation. It’s important to note that this document does NOT reflect redactions related to the creation of the General Release dataset. Extensive numbers of variables or whole modules were removed from the Restricted Release to create the General Release. So the document is really only applicable to the Restricted Release data. NSCAW 1 and NSCAW 2: How the data were collected. The questions making up the NSCAW study, were administered via a computer using the Audio Computer-Assisted Self Interview, otherwise known as the ACASI software. The ACASI system allowed for the implementation of logic for qualifying to receive a module, skip patterns between questions within a module, and custom wording based on preloaded information like names, dates, ages, etc. The teacher and local agency director interview were administered via paper and pencil. 
NSCAW 1 Other Sources of Information. NSCAW 1 also has data originating from the 1990 US Census, the National Death Index, and they obtained salivary cortisol measures which were collected from the infant sample at wave 5. 
NSCAW 1: CPS and Long Term Foster Care Notes. The sampling weights contained within each of the CPS and Long Term Foster Care data files are specific to that sample. Under NO circumstances should the two samples be combined and analyzed in any weighted analyses. I feel like I field that question quite often and just wanted to make sure that everyone attending this session is aware that you can’t combine the two. It would be difficult to imagine any circumstance, even unweighted, where it would make sense to combine the two samples. And I do know I think there’s like one or two publications out there where they in fact did that and I would say that it was probably early on and when NSCAW was released and perhaps not a lot of knowledge was known about combining those two samples. So NSCAW 2, the study largely mirrored the NSCAW 1 study with the following differences. Of course they sought to examine changes in the 9 years since NSCAW 1. NSCAW 2 had 5,872 children from birth to 17-and-a-half years old at time of sampling. So the NSCAW 2 sample is starting out a little bit older than the NSCAW 1 sample because NSCAW 1 was like capped at 14. They were sampled from CPS investigations that closed during a 15-month period beginning February 2008. Only 3 waves of data collection, and not the 5 that we had with NSCAW 1. 76% of the participating counties from NSCAW 1 were retained for NSCAW 2. There was no separate Long Term Foster Care sample for NSCAW 2.
Standardized measures were updated to reflect updates rolled out by measure creators. There was no 12-month telephone follow-up interview which would have corresponded with the Wave 2 data collection from NSCAW 1.
There was no Teacher Survey at the 36-month follow-up which is Wave 3. In NSCAW 2 we have linked administrative data from the National Child Abuse and Neglect Data System also called the NCANDS Child File and the Adoption and Foster Care Reporting System also called AFCARS. We also house those two datasets and many other summer training webinar series have been devoted to teaching folks about those two datasets.
NSCAW 2: Data collection timing.
This table shows the dates of data collection and the span of time between waves of data collection. The top row of information contains the data collection start and end dates for each of the waves. The bottom row shows the number of months after the close of investigation to a specific data collection wave. Wave one was collected approximately 2 to 10 months after the close of the CPS investigation that qualified the child for inclusion in the study, 18 months for wave 2, and approximately 36 months for wave 3, which varied depending on the age cohort. 
NSCAW 1 and 2: Statistical Weighting. If you’re not already glazed over from all of the sampling design talk then the statistical weighting might be send you making you feel a little bit more overwhelmed. Just stay with me and we’ll get through it.
So in order to obtain unbiased estimates, the complex sampling design must be accounted for in your analyses through the use of statistical weighting variables included in the data files. The three main weight variables are the National weight which is the NANALWT variable and there’s a national weight for each wave because remember we’re adjusting for non-response at each wave. You have a STRATUM variable, and you have a primary sampling unit variable. Analyzing the data without applying the appropriate weight or any weight can lead to misleading results. Analyzing the data without the weights means that the NSCAW dataset is treated like a sample of convenience rather than a nationally representative dataset. So you lose a little bit there.
The restricted data file contains the Stratum and NSCAWPSU variables which are needed to obtain standard errors that are correct for the NSCAW sample design and for this reason we strongly recommend that analysts use the Restricted Release version of the data for any analyses intended for peer-reviewed publication.
The wave 1 weight reflects the selection of probability for participation in NSCAW, while adjusting for nonresponse, and under-coverage, which refers to the sampling frame and is not covered by this presentation. This information is described, in detail, in the Data File User’s Manual which is the equivalent to what we would call our user’s guide. So if you’re interested in learning more about that we can certainly share the documentation with you.
After wave 1, each weight is adjusted at each wave for wave-based non-response.
Working with a weighted dataset can be intimidating. Please know that like we are here to answer your questions and help guide you along the way. So don’t let the weighting deter you from potentially using this dataset.
So, more about NSCAW 1 and NSCAW 2 Statistical Weighting. There are additional weights available for specialized examinations of the data. These weights are available to Restricted Release data users by request. We have Agency weights which are used for multi-level modeling when analyzing children nested within an agency in two-level models and time nested within children nested within agencies for three-level models.
We also have Calibration weights. They allow for comparisons between NSCAW 1 CPS sample and the NSCAW 2.
The weights are adjusted to account for the differences in the target populations for NSCAW 1 and NSCAW 2.
For more information see the video presentation entitled, “Comparing NSCAW 1 and NSCAW 2 using Comparison Weights” found on the NSCAW User Support page of the NDACAN website. 
The calibration weights are largely under-used. I don’t know of any publication from our data users that has leveraged these weights. So if you’re looking for an area that really has not been well-explored, that would definitely be an avenue to examine.
NSCAW 1 and 2: how are the data organized? 
The data are oriented wide, so they’re one-record-per-child, with each wave’s data contained within its own data file.
NSCAW 1 and NSCAW 2 each contain approximately 40,000 variables. Of course you don’t need all 40,000 to answer the research questions that you have, but you’re going to have to be well-versed or willing to learn how to pare down those files and merge them together.
There is an across waves data file containing summary variables based upon variables administered across multiple waves of data collection or that are generally applicable to all data collection waves like the Census variables and information from the National Death Index.
An example of information that would be summarized across all of the waves of data would be the Child’s Living Environment module, which is administered to the current caregiver in the form of two different modules LE and LV. One represents the baseline and one represents follow-up. And the caseworker was also asked the same information and that’s in module LN. 
The module captures changes in the child’s living situation or what we would call placements. Data from the module is very complex and challenging to use as there are multiple informants and multiple waves of data. In addition, the Living Environments modules contain a looping sequence in order to capture all living situation changes. This means that when a respondent answers “yes” there was a living environment change, then the sequence of questioning about that change begins again looping through each change until there are no more changes. So you have layers and layers and layers of information there.
The Across All Waves data file helps to organize that information and also provides summary variables for commonly sought--after derivations like the “Cumulative number of days in out-of-home care” or the total number of you know placement changes, things of that sort is what you will find in that Across Waves file.
The Advantages of NSCAW
For the CPS samples: When the weighting is applied correctly, your results are generalizable to all children in the U.S. who are subjects of child abuse or neglect investigations or assessments conducted by CPS and who live in states not requiring agency first contact. So that’s pretty remarkable that we can we can make those generalizations at a national level. In NSCAW I, you have the added ability to generate state-level estimates for the 8 largest states that are identified in the dataset. To do this you would apply the state weight instead of the national weight, and those weights are also included in the dataset package that gets delivered when you order the data.
There are multiple informants and with multiple informants when you ask the same question to you know three different people you may get different answers. And so RTI International did a great job of creating derived variables that sought to resolve those discrepancies across informants. 
So you have a collection of derived variables at each wave that you can leverage if you agree with how they resolved those discrepancies so that you know you kind of don’t have to do that for yourself. You can just use what’s there. 
It is very well-documented. You have an HTML codebook that just you know you click on a variable name and you can see frequencies. You can choose whether you see unweighted frequencies or weighted frequencies. They have the data file user’s manual which talks about the entire process for collecting the data, how to apply the weights, like everything. And you have an appendix that has every question that was asked of participants as long as the the measure itself wasn’t copyright protected or if they weren’t able to secure permissions to include it then it wouldn’t appear in that document but it’s really nice to be able to see how the question was asked to participants. Also in that document is the logic for having received the module and also any skip logic that occurs between the questions. Which can be really instrumental in tracking down missing data.
The statistical weighting and size of the dataset, both the number of variables and the amount of documentation to read, can feel like way overwhelming. There is a steep learning curve that comes with using NSCAW but it’s it’s worth learning because there are not many child welfare datasets out there that will allow a secondary analyst to make generalizations at a national level. Learning either NSCAW 1 or NSCAW 2 will have general applicability to the other dataset and likely to any future NSCAW study. It’s 100% worth your time and energy investment and just know that I’m here to answer your questions along the way. So don’t don’t shy away from the dataset because you feel like it’s like a lot to take on it’s it’s definitely worth it you can really just examine almost anything child-welfare related.
To examine what has already been done using the NSCAW datasets please visit our NSCAW collection of bibliographic citations captured in our online searchable Zotero database called the child abuse and neglect Digital Library or canDL which is located on the “candL bibliography” page of our website. One of the other questions I often field about NSCAW because it has been released for a number of years now.
Don’t assume that because these datasets have been available for a while that they are somehow “used up” or no longer relevant. There are approximately 40,000 variables in each of those datasets. I’m confident that there are still many facets of those datasets that have yet to be explored. So don’t count that out.
OK now we’re going to switch gears and we’re going to talk about the Longitudinal Studies on Child Abuse and Neglect or LONGSCAN.

What is LONGSCAN?
LONGSCAN is a consortium of research studies operating under common by-laws and procedures. It was initiated in 1991 with grants from the National Center on Child Abuse and Neglect to a coordinating center at the University of North Carolina at Chapel Hill and five data collection sites. Each site conducted a separate and unique research project on the etiology and impact of child maltreatment. 
The goal of LONGSCAN was to follow 1,300+ children and their families until the children themselves become young adults.
The LONGSCAN Investigators included Desmond Runyan, Howard Dubowitz, Diana English, Jonathan Kotch, Alan Litrownik, Richard Thompson, and what we refer to as the LONGSCAN Investigator Group which is all of the folks that worked on the project: collecting the data, organizing the data for submission for archiving. They truly did an excellent job.
The LONGSCAN Sponsors. Funding for the LONGSCAN study was provided by the U.S. Department of Health and Human Services, Office of Child Abuse and Neglect.
The early years of data collection were sponsored by the National Center on Child Abuse and Neglect under the Office of Human Services until NCCAN became a part of OCAN in 1998.
OK a little bit about the LONGSCAN Sites. 
The East (EA) site is considered to be urban. 282 children selected from clients of three pediatric clinics serving low-income, inner city children. There are two risk groups. There’s the Child-based risk group which is inadequate growth in the first two years of life. And then there’s the parent’s group of having an HIV infection or drug use. The South (SO) is a combination of urban, suburban, and rural. 243 children were drawn from a population of children identified as high risk at birth by a state public health tracking effort. Non-reported children were matched to reported children in a 2-to-1 ratio. The Midwest (MW) site was largely urban. 245 children of which two-thirds were recruited from families reported to CPS with half receiving comprehensive services and half receiving CPS intervention only. The other third consisted of neighborhood controls.
The Southwest (SW) site was suburban. 330 maltreated children who entered a county dependency system due to confirmed maltreatment. All children were in an out-of-home placement with a relative or foster family. 
The Northwest (NW) site was urban, had 254 children who were judged to be at moderate risk following a report to CPS for suspected child maltreatment. Approximately 60% of the referrals were later substantiated.
LONGCAN Data Collection. So LONGCAN data collection began in July 1991 and went through January 2012. In general, a visit corresponds with a child’s approximate age at the time of interview. Comprehensive in-person assessments occurred approximately every two years at Visit 4, 6, 8, 12, 14, 16, and 18.
Annual Contact Interviews were conducted by phone and occurred on years between the comprehensive interviews. So we have visit 7, 9, 10, 11, 13, 15, and 17. There are multiple informants: there’s the Children, their caregivers, and teachers. Data from CPS reports were collected via record abstractions where actual study staff went into the agencies and were allowed to review the contents of the agency file. I guess it was like parts of it, it wasn’t the whole thing. The total number of children who participated in the study was 1,354 across all 5 of the sites.
The LONGSCAN Comprehensive Assessments. Assessments were interviewer-administered for visits 4, 6, and 8. At age 8 they used computer-assisted face-to-face interviews. Age 12 onward, interviews were administered using the Audio Computer Assisted Self-Interview the ACASI system which we previously heard about in the NSCAW study. Teachers completed mailed paper-and-pencil forms beginning at age 6. Data were collected using approximately 144 different measures.
These are a mix of commonly used standardized measures and some of them were project-created measures and some of those project-created measures were then used in NSCAW. And you’ll see that if you get the documentation, you’ll see those references.
And I’m going to have my co-host post a link to the Appendix A: Measures Administration Schedule by Site and Age document in the chat window. The document shows the full list of measures, the sites in which they were administered and the visit or age. The document is also available for download from the LONGSCAN dataset page of the NDACAN website. And I also would like to say that the sites all were kind of kicked off their data collection at different times before joining the LONGSCAN consortium so around age 8 is when you’ll see more consistency in the measures being administered at all of the sites. So when you’re examining that document that’s what you can kind of expect to see.
LONGSCAN Data Organization. The information collected from the measures and CPS record abstractions are represented across 351 data files.
Typically, the item-level data are contained in one data file and the scored data are in a separate data file.
Data files vary in structure from being one record per child ID to being multiple records per child ID where a record represents a child-visit pair. This means we are talking about some files being wide and some being long or stacked. This will require some experience or willingness to learn how to restructure or reshape the data to get it into the form in that you need it for your specific analyses and also you’re going to have to do a lot of merging of those files.
The advantages of using LONGSCAN. It allows for longitudinal analysis with its many waves or time points of data collection.
The data were collected from multiple informants to measure outcomes and intervening factors that may influence the link between risk status and outcome.
The study used both standardized and project-developed measures with psychometric properties being available for nearly all of them in the Measures Manual.
The dataset is well-documented with Data Dictionaries that often provide the full text of the question asked to participants. The Measures Manuals provide detailed information regarding the development of the measures or modifications to the measure when applicable. 
Statistically speaking, it’s more straightforward than examining a weighted dataset.
To examine what research has already been done using the LONGSCAN data, please take a look at our online searchable database of bibliographic citations called the child abuse and neglect Digital Library or canDL. You can access the canDL by visiting the “canDL Bibliography” page of the NDACAN website. 
And now we’ll shift gears again and we’re going to talk about the National Child Welfare Workforce Initiative Comprehensive Organizational Health Assessment 2 or “NCWWI”. 
What is NCWWI? The National Child Welfare Workforce Institute is funded by the Children's Bureau to increase child welfare practice effectiveness through workforce systems development, organizational interventions, and change leadership. The archived data come from the NCWWI Workforce Excellence initiative. As part of the initiative, child welfare staff from three sites completed a baseline and follow-up Comprehensive Organizational Health Assessment to identify critical workforce strengths and challenges. 
The NCWWI Investigators are from the Butler Institute for Families at the University of Denver. And I apologize if I mispronounce anyone’s name. We have Robin Leake, Shauna Rienks, Anna de Guzman, Amy Hewho is also a former Summer Research Institute participant, Mary Jo Stahlschmidt.
The NCWWI sample and study design. Participants were child welfare staff employed at one of three public child welfare agencies that were a part of the NCWWI Excellence project.
Only participants who agreed to have their data used for research purposes are included in the archived dataset. 
The Comprehensive Organizational Health Assessment was administered at two time points via an online Qualtrics Survey. An entire agency was canvassed at each administration of the assessment. This means that if someone was working at the agency at baseline and no longer working at follow-up they would only have one time point. Likewise if someone was not working at the agency at baseline but was hired prior to the follow-up assessment, they would have been solicited for participation in the study and would have a record for only the follow-up time point. This results in 2,832 child welfare staff completing the baseline assessment and 2,912 staff completed the follow-up assessment, with 1,034 staff completing both administrations of the assessment. 
The NCWWI Data Collection Sites. We of course have to mask the names the specific locations of these sites to protect the confidentiality of the participants. So we have two mid-western state-administered child welfare agencies and one west coast urban county-administered child welfare agency.
The NCWWI Constructs Measured. I didn’t have a document ready for this one so we’ll just go through the list really quick. They measured Burnout, Satisfaction with relationships with community providers, Coping strategies, Exposure to violence, Inclusivity, Intent to stay, Inter-professional collaboration, Job satisfaction, Job stress, Supervision (for both supervisors, managers, and frontline staff), Shared vision, Traumatic stress scale, Perceptions of child welfare, Leadership, Learning culture, Professional development & preparation for work, Physical environment, Professional sharing and support, Organizational climate, Readiness for change, Self-efficacy, Organizational bias, Peer support, Workplace prejudice and discrimination, Time pressure, and Team cohesion.
The NCWWI Data Organization. 
There is only one data file containing 4,710 records. A record represents a child welfare staff at a given time point. Please keep in mind that any numbers reported during this presentation for the NCWWI dataset are preliminary and are subject to change before the official release of the dataset. Because this is we’re just giving you a sneak peek here.
The advantages of using the NCWWI Dataset once it is released. These data are being made available to secondary analysts in the child maltreatment community for the first time very soon!
This means that there will be lots of unexplored territory. 
And these data represent an in-depth examination of the experiences of child welfare staff working in public child welfare agencies which really has not been captured in other datasets housed here at NDACAN.
OK so some important points before I turn things back over. Please remember that some of the datasets discussed today are currently in the field collecting the next iteration of data, therefore we should take care in how we frame our discussions to avoid influencing participation in current or future studies with the same or similar names.
Detailed questions not suitable for the webinar or requests for study documentation can be directed to NDACANsupport@cornell.edu.
When submitting requests to receive the documentation, please be sure to include your full name, contact information, and institutional affiliation.
Thank you for attending today’s session, please feel free to type your questions into the Q and A portion of the chat window. And if by chance there’s so many questions that we can’t get to them all, please feel free to submit your questions to that email address that I mentioned a minute ago and I’ll send a response in the next 24-48 hours.  
OK so I think it’s back to you Erin.

[Erin McCAuley]
 Yeah thank you so much Holly that was fabulous. I know that's like a ton of content and so that was a really great overview and especially exciting preview of the new information coming. Clayton Covington who is our research associate is going to help moderate the Q and A so if you have questions please put them into that Q and A box, Clayton will read them out loud and then Holly will do her best to answer them.

[Clayton Covington]
Alright thank you for that Erin we're going to start with the Q and A. Our first question for you Holly says “how do these datasets work in SPSS, specifically how do you compare across waves in SPSS?”

[Holly Larrabee]
 I'm going to speculate that they are asking about the NSCAW so for NSCAW SPSS has a module I believe that you can add on to your SPSS license called The Complex Samples Module, and that can appropriately handle the sampling design for NSCAW and apply the weighting correctly. And once you have that module you just run your analyses through through that portion of SPSS. So you would just go into the complex sample module and you would say you know run a frequencies and it would once you specify your weighting variables and everything would SPSS would account for the weighting correctly.

[Clayton Covington]
Alright thank you for that Holly. Okay looks like a follow-up perhaps to that last question because it's from the same participant, says “Also, 1. Is procedural justice measured? And 2. how much would a person at a nonprofit/ national organization get access to these datasets without a faculty sponsor, like if you are the only data person at your organization?”

[Holly Larrabee]
 So I don't know if procedural justice was measured. I think the best thing to do would be to take a look at the construct documents that were the links were posted to, and also you know potentially requesting the study documentation. I think I'm going to have Andres answer the data ordering question regarding access to the datasets for nonprofits without like a faculty sponsor

[Andres Arroyo]
 Yes you would go through the regular order process and there's a spot on the form to describe your situation and each application gets reviewed and you can provide supplemental documentation about how your organization handles sensitive data and so forth.

[Erin McCAuley]
Thank you Andres yeah so we you don't need a faculty sponsor to get the data and we're you know we're really happy to work with organizations to access the data and kind of meet whatever hoops are required for different datasets. So you can reach out to us through the website or the support line but you can also just go ahead and fill out the order forms to the best of your ability and we'll work with you to get it approved. But yeah we have many nonprofits that use our data and are not affiliated with a faculty member.

[Clayton Covington]
Okay next question. Can you briefly reiterate if there are any differences in source populations between NSCAW and LONGSCAN?

[Holly Larrabee]
So they're two different datasets and they pulled their they pulled their samples from different...they're separate so you would only analyze LONGSCAN data that's contained within LONGSCAN and you would only analyze NSCAW data that's contained in NSCAW. The LONGSCAN samples were considered to be at-risk samples. NSCAW, the samples had contact with the CPS they had a CPS investigation. So in that regard it's like LONGSCAN is more like at-risk, maybe didn't have contact with CPS just yet, but NSCAW had to have contact with CPS in order to be eligible for the study.

[Clayton Covington]
Alright thank you Holly. Next question asks, “Is there weighting system for comparing/ combining samples across the five LONGSCAN sites?”

[Holly Larrabee]
So there's not a weighting system but I think the approach that the LONGSCAN investigators take is to include site in your model. So whatever model you're trying to run, including site as a covariate is one of the strategies that they've used. And if you take a look at the child abuse and neglect Digital Library you can examine some of the publications that the LONGSCAN investigators the original investigators published and get a sense for how they went about using data from all five sites in their analyses.

[Erin McCAuley]
Yes and I just popped the link for canDL into the chat. So that's definitely the best first place to start.

[Clayton Covington]
“So for the NCWWI data, what has been the timeline for data collection?”

[Holly Larrabee]
You know I don't know that dates right off the top my head. I think it started being funded in 2008. The one that we're receiving is called, it's referenced as 2, so it makes me feel there was data collection effort before the one that was archived. So I'm just not it's not it's not right on the top of my head what the timeline is but I can definitely follow-up afterwards with the timeline.

[Clayton Covington]
Okay, and so also related to the NCWWI dataset, “when it mentions staff does it include agency directors/ supervisors and if yes, is there a way to know if certain supervisor respondents were connected to caseworker respondents?”

[Holly Larrabee]
So yes they did interview agency directors and supervisors. The second part, is I am not sure if you're going to be able to identify, I'm trying to think, I don't think you're going to be able to identify like match the supervisor to the subordinate respondents. But it could be possible just based on you know if they are all within the same site you might be able to make that connection but like right off the top of my head I don't know if that's possible.

[Clayton Covington]
Okay so the next question seems to be related to an earlier question. And again this person did not specify which datasets they were referencing but I am assuming it's the NSCAW. And it's a follow-up on the meaning of procedural justice from earlier and they clarify “procedural justice meaning perceptions of courts, attorneys, and social workers, or views of the legitimacy of the system, or even satisfaction within the system.” And I think they're asking whether or not that would be, if any of those components are measured within the NSCAW dataset.

[Holly Larrabee]
No, so I would say that the local agency director interview might be the closest thing that you would come to in NSCAW. And you know again that's the CPS agency director. They may have some measure of satisfaction with the system but definitely you know, no, I don't believe there any assessments or anyone was asked about the courts or attorneys. There's parental I believe there's caregiver reports of like their satisfaction with the caseworkers that were working on their case, but I don't recall there being anything about courts or attorneys.

[Clayton Covington]
Okay. We also have another question: “Are the NCWWI samples representative of all child welfare workforce of the three states that are sampled?”

[Holly Larrabee]
 I would say the two Midwestern states would just be representative of the two Midwestern states. And the West Coast urban location you know again we can't disclose those exact locations because that would pose a threat to the participants' confidentiality. But yeah I think you know the Midwestern states you could definitely make generalizations about it you know being a Midwestern state. I don't know it's not the same as like NSCAW if that's kind of what you're asking, like it's not there's no weighting to apply to make it represent like state-representative or nationally representative. But just generically speaking I believe that they attempted to canvas... I'll have to examine that closer but I think they attempted to canvas a number of agencies within the two Midwestern states. I mean I don't know if it was all of them.

[Clayton Covington]
 okay thank you Holly.

[Erin McCAuley]
Well we are right at time. So I just want to thank Holly again. That was an incredible presentation. I know this is something that participants have been asking for so it was really great to have session. And again next week we're going to be talking about admin data and how to link it. At the end of the summer you'll receive an email from me with a survey about the summer training series so if you get that please fill it out. It's how we improve the presentations over the years and it's also how we source ideas for what you guys like to hear about in the summer sessions for example we had this session because a few participants last year said they were really missing information on the survey-based data having used that administrative data before. And so we were able to incorporate it into this year's summer training series. So thank you again to Holly, Clayton, and Andres and we hope to see you all next week!

[Voiceover]
The National Data Archive on Child Abuse Neglect is a collaboration between Cornell University and Duke University. Funding for NDACAN is provided by the Children's Bureau, An office of the Administration for Children and Families.