[Voiceover] National Data Archive on Child Abuse and Neglect. [Clayton Covington] alright good afternoon everyone and welcome again to the National Data Archive on Child Abuse and Neglect summer training webinar series. All of the sessions for the summer training webinar series will be recorded and will be transcribed and made available on our website at a later time. You can also view previous years'sessions on our website but also just a few guidelines in terms of questions. So you all should see in your screens and ability to ask questions via the Q and A section. We ask that you direct all your questions there and we will forth questions until the end so feel free to ask them as they come and put them in there and then I will read them at the end of the presentation and our presenter Alex will be able to go over them. So again we are the National Data Archive on Child Abuse and Neglect posted at the Bronfenbrenner Center for Translational Research at Cornell University but if you participated in our last week's session you'll know that we are also now affiliated with Duke University as our Co-PI and current codirector Dr. Christopher Wildeman is moving his affiliation with Duke University so number of our team including our codirector and Co-PI are going to be at Duke University. You can take it away Alex. [Alex Roehrkasse] okay my name is Alex Roehrkasse I am a postdoc at NDACAN I'm also one of these people who is going to be transitioning from Cornell to Duke this year. I joined the Archive back in August. I'm trained as a sociologist but a lot of the work I do at the Archive is building out our holdings of historical data and so I'm going to be leading the training series today and next week both of which will focus on historical data. Today I'm going to talk mostly about our efforts to sort of collect and harmonize historical data and then next week I'll offer sort of an illustration of some research you can do with those kinds of data. So I'm sorry not to be able to see any of you or you to be able to see me but I'm excited to give this presentation. I hope it's interesting and helpful. Okay so let's get started. I think that's the theme for this year's training series is New Horizons for Child Welfare Data and so I think we are thinking about historical data acquisition has sort of one of these new horizons. Like I said today session is going to focus on some of the existing and new historical data that researchers might be interested in using and then next week will be a research illustration using some of those data. And so today the talk is going to kind of be broken down into three parts. So first I'll talk about why anyone should be interested in doing historical research on child welfare and what that research might look like. Then I'll talk about you know if you wanted to to historical research on child welfare how would you actually use archive data to do that. And then lastly I'll basically kind of offer some descriptive analyses of some historical data just to kind of illustrate from the outside with kinds of things we're able to show if only from a descriptive sense. Okay so why should we use historical data? Or maybe you can just take a step back from that question but what is historical data? So what is this presentation going to be about? In some sense all data is historical data because it describes sort of by necessity things that happened in the past. So for example you could think about the National Incidence Study has historical data because it's now old. That's not really what I'm going to be talking about today. Today I mostly going to be talking about data that describes attributes or events or processes that unfold over historical time. And by historical time I mean sort of time with respect to larger groups or even society. So for that reason there are a lot of kinds of data that have to do with time, timeseries data, annual data, survival data that nevertheless sort of pertain to the individual level. And that's not the kind of thing we're going to be talking about today. Mostly were going to be talking about relationship between time and units of analysis that are larger and for that reason we are also going to be talking about data that's aggregated. So a lot of the data in question starts at the individual level. The data-generating process is sort of an individual-level process but then we'll be aggregating that data to larger geographic units of analysis whether it's the County or the state or the entire United States and then tracing variation and change and certain kinds of outcomes of interest at those larger aggregated geographic levels. So why would one be interested in doing historical analysis on using child welfare data? I think most straightforwardly we might be interested in documenting change in historical time. So if we were to observe for example rates of child neglect in the United States in 2018 and we saw that they were a certain level and we thought that level was high. We wouldn't really have a basis for saying it was high without having some context for those levels. And so by referring to historical data we can understand the present a little better, including recent trends. So for example if we only had data for 2017 and 2018 and we saw that rates of child neglect went up, we might be curious whether or not that increase over one year was the continuation of a longer-term trend or an anomaly in a larger trend that was actually downward sloping. Similarly to the way that levels of certain events might change over time, the relationship between things might change over time. So for example, if we observed a strong association between child poverty rates and rates of child neglect, we might furthermore be interested in whether the relationship between poverty and neglect was stable across historical time or whether those things used to be more strongly associated and are now sort of becoming decoupled, or perhaps the opposite that child poverty is becoming over time a stronger and stronger predictor of child neglect. Historical data can also be really helpful for doing causal analyses. So if we're interested in the effect the causal effect of one factor on some outcome of interest relating to child welfare, having historical data can be very helpful especially when we have observations over time for a particular geographic unit of analysis. So we have multiple observations, counties or states over historical time we're able to control for a lot of observable but also unobservable things that help us to increase the certainty with which we can identify so causal effects of things for example welfare generosity on child neglect. And then lastly if we are interested in protecting the future, having information about the past is quite helpful. So I'm working for example right now with a team that's trying to develop forecasting models for rates of children entering into substitute care and so were using a lot of the historical data I'll be describing today to build models that help us anticipate state-level changes in foster care entrances. Okay so let's say you've decided you want to to some historical research, you think you have a research question that is well-suited to explore using historical data. What kinds of data at the Archive would be helpful for doing that kind of research? The data I'm going to talk about today and next week are all the administrative data. So the Archive does hold a good amount of survey data, and those data can be used for historical analysis but there's an overwhelming advantage using administrative data because they tend to be more easily comparable across time and you don't have to deal with a lot of sort of sampling issues that can cause problems especially over time. So we're going to be talking about the administrative data today as a kind of fall into two groups. First administrative data that's already available on the Archive that you could access today and kind of start a historical project. And then new administrative data that's not yet available to researchers but should become available within the next year. And so researchers are interested in planning historical projects should be aware that some of these data are going to come online. So the two existing data sets that are most helpful for doing historical analyses are the AFCARS which captures children entering into and in foster care and then also adoptions. And then the NCANDS which captures incidences of abuse and neglect. Both of those data sets have quite a wide historical range starting in the early to mid-1990s and with a few exceptions that I'll note, they are fairly consistent across time. And so it makes it easy to do sort of long term historical analyses. Over the last year the staff at the Archive have been working to develop to other data sets that are predecessors to the AFCARS. That Voluntary Cooperative Information System was sort of an explicit predecessor to the AFCARS. The AFCARS was in some sense modeled off of the VCIS. And it is similarly interested in children in the foster care system. The Children's Bureau Statistical Series some of you may be familiar with, this was a long-running series of publications from the Children's Bureau that covered a wide range of topics including sort of personality and child welfare, juvenile justice statistics, but one of the things they published was children receiving welfare services by living arrangement. And so we can with a little bit of creative manipulation create time series which we then link to both the VCIS as the AFCARS creating timeseries data on children in foster care that ranges from the early 1960s through more or less today. So to give a little more detail about these new sources of historical administrative data, the VCIS was commissioned by the federal government but administered by the American Public Welfare Association and their survey instruments asked states to report aggregate data. So there are no individual-level data in the VCIS. We only have aggregate data. And so those data are annual state-level data on children entering into, residing in custody, and exiting substitute care each year. And those aggregate data are cross-tabulated by race and ethnicity, by age, by sex, by living arrangement in foster care whether they are in foster care home or an institution or group home, independent living. Those data aren't tabulated by multiple attributes though, so it's not possible from within the VCIS to get for example counts of children in foster care by race, ethnicity and age. The Children's Bureau Statistical Series is more limited and information we have. So those are just counts of children in substitute care in each state in each year. So there's no demographic information in a CBSS. And so these data are great because they allow us to sort of combine different sources to create really long time series but they do have limitations that are really important to know. The first is reliability. For example because the VCIS was voluntary, states exhibited quite a bit of variation and how they understood the meaning of children in substitute care. And so in any given year states might be sampling or reporting information about slightly different populations of children. And so we need to take great care in making specific comparisons across states or within states across time. Moreover, none of the sources I'm going to talk about today have complete data. So we don't have that much information about the degree to which states might not reports children in years they were reporting any children. But we do know, of course, years in which states neglected to report any information to the collecting agency. So for example as you'll see there are some years where each source contains information from only a subset of the states. And if we're not careful about how we deal with those missing data, some of our estimates can be biased. And so a lot of what I'll talk about going forward is how we might think about dealing with missing data. Okay so let's say we want to do an historical analysis and we've even identified which archive data will be helpful for doing that analysis. Okay how do we actually go about doing it. So it's important to note that some data is already aggregate but other data products are only available through the Archive at the individual level and researchers who are interested in doing historical analyses would need to aggregate them themselves. So the NCANDS comes in two different forms: the Child File and the Agency File. The NCANDS Agency File aggregates data already for users to the state level. So if someone was interested in doing a state level analysis of abuse and neglect you can download the NCANDS agency file and it would be ready to go. Similarly as I just described the VCIS and the CBSS are already aggregated. The AFCARS in the NCANDS Child File, however, are available to researchers only at the individual level. And this is both a blessing and a curse. It takes a little more work to manipulate the data to get them into aggregate form but it also gives researchers much more flexibility in how they aggregate the data. So if you are interested in specific subpopulations, for example, you can aggregate those subpopulations with much greater flexibility than is available in the other pre-aggregated data sources. The AFCARS Foster Care File and the NCANDS Child File include IDs for not only states but also counties so it's possible to do county-level analyses with those data sets. The Adoption File doesn't include County information so you can only do state-level analyses with the AFCARS Adoption File. Okay I won't talk about this at great lengths but I've thrown three slides and here for folks who haven't dealt with aggregating data before. Depending on which software package sure most comfortable with here are just some little snippets of code that illustrate how you would collapse individual-level data to any level of geographic aggregation that you are interested in. So for example if you are a Stata user and you have individual level data, and you have a variable ID that is unique for each individual child, you have state variable that was a unique identifier for each state, a year variable that was unique for each year you could use the collapse command and then count the number of individual children you had, and then the "by" clause specifies the variables which jointly identify your unit of analysis. So if you wanted counts of children by state year, you would collapse by state year. If you wanted to count children by race you would collapse by state and year and race, and you'd be left with counts of children in each state in each year in each racial group. The logic is very similar if you're working in R especially in the tidyverse, just the sequence is a little bit is the other way around. You first group your data by the aggregate identifiers, state, year, or state year race. And to be clear you could include any number of identifiers you could tabulate furthermore by sex or age. And then instead of collapse you would use the "summarize"command to count the number of children in each of those groups. And then I am not an SPSS user but here's just a brief illustration of how you do essentially the same process in SPSS. Okay so let's say we either downloaded aggregate data or we aggregated our own individual level data what we're left with is a count of children. Counts can be interesting and useful in their own right but often they don't actually tell us what we want to know. And that's because the underlying populations for which these counts are derived are of varying sizes. And so very often what we are actually interested in is a rate or a proportion and so we need some denominator to make that count meaningful. Sometimes archive data themselves provide a useful denominator. For example let's say we were interested in the proportion of children entering into foster care in each year who were American Indian or Alaska Native. Then we would take the count of American Indian and Alaska native and we would divided by the count of all children entering into care. And so the Archive data provide both the numerator and the denominator for the proportional variable that we are interested in. Other times so if were interested in population great we need population level information to construct that rate. And so where you would get that data is going to depend a little bit on what your specific research question is but I've just enumerated here a few different resources that are very commonly used to construct rates. So for example the SEER files from the National Cancer Institute harmonize the Census Bureau's small area intercensal small area estimates so you can get annual data that County level by the age and race/ethnicity. You can then aggregate those County data to the state level if you are interested in a state-level analysis. The NHGIS is a great platform for harmonizing Decennial Census data and the American Community Survey. And then of course if you were doing analyses on a specific subpopulations to might not be able to find aggregate data about the population that would be specific enough to construct the exact rates or proportions you were interested in. In that case you could do the same thing for your denominator as you did for your numerator: take individual level data and collapse particular groups of those individual level data to whatever aggregate geographic unit you are interested in. So for example you could use the IPUMS data whether from the Decennial Census, the ACS or the current population census and collapse those data to create your denominator. So you may start to be picking up on a pattern here. I sort of promise a lot of exciting opportunities with historical research and then I tell you to be very very careful so here are a few more cautionary notes you should think of as you're moving forward with any historical research. The first is to ask yourself are my data consistent across time? I said before that one of the great assets of historical data is they are relatively consistent. That's true within reason. There are important changes that happen across years in any given source. So for example there were really important changes in the way race and ethnicity were measured in federal government statistics around the turn-of-the-century. And so for example if you were doing analyses by race/ethnicity you would want to take care in making any comparisons before and after the year 2000. Moreover you would want to think carefully about whether your data are consistent across the different sources. So for example I said that the AFCARS was modeled closely off of the VCIS and so they are generally fairly comparable. That's true but there are important differences. For example the AFCARS includes information about children who have run away from their placement. Those children don't show up in the VCIS or the Children's Bureau Statistical Series. That's very small population and so it's maybe a compromise that were willing to make as a researcher but it's important to know that we are making that compromise. Lastly you'll want to think carefully about how missingness affects your data, how suppression of information by the Archive affects your analyses, or issues of sampling in other data combined with archive data might affect your research. So for example as I said before there are some years where certain states don't report any data. That could be an important source of bias in your analysis if you're not considering the process that leads states not to report information. The archive suppresses certain geographic information that it thinks might lead to the disclosure of private information, basically allow the identification of children in the data. So for example both the AFCARS and the NCANDS suppress County identifiers for all individuals coming from counties containing fewer than 1000 observations in that County. So the pression is going to change from year to year depending on how many children do or don't show up in the sources and so you wouldn't want to just you wouldn't want to do nothing about that missingness. You'd want to sort of be intentional about how you dealt with that suppression. And then lastly if you were to use survey-based data to construct denominators for example that would introduce sampling error into your rates or proportions. So that sampling error wouldn't come from the administrative data from the Archive, it would come from other sources but you nevertheless needs to deal with and report that sampling error, that uncertainty deriving from other sources. Okay so all that said, you've settled on a historical question, you've identified the historical data that you want to use, you've even downloaded it, manipulated it to get it ready, what would it look like then when you got all that data together what kinds of things could you learn? Well this figure illustrates kind of if you downloaded all these sources and you combined them, how complete a picture would you be getting? So the horizontal axis here plots historical time and vertical axis plots the number of states in any given year which report data to the reporting agency for any source. So 51 means we have complete data for the entire U.S. that's 50 states and the District of Columbia. And then whenever a dot falls below 51 it means we have fewer than all states reporting in any given year. And then the color of the dots indicates the source of the information in that year. And so you can see for much of the 1960s from Children's Bureau Statistical Series we have complete data on children in substitute care. In the 1970s though, there are a number of years where we either don't have any sources reporting any information about children, but we have spotty information sort of Children's Bureau Statistical Series system that's kind of in decay. You'll see a pattern that whenever a new source is coming online or starting to decay it's when we started to lose a number of states from the analysis. In one year you'll see in 1995 we have information from both the VCIS and the AFCARS. The figures may be a little misleading. We have 51 observations for the VCIS in that year but I think only 16 or 17 observations from the AFCARS. That's going to be really important though for us as I'll talk about in the next slide in evaluating the reliability of our data. And then thank goodness after 2000 once the AFCARS is sort of well-institutionalized we have every state now reporting data consistently for the 21st century. So this figure I mentioned just immediately previously that we have data from 1995 from two different sources, this figure helps us leverage that opportunity for comparison to evaluate the reliability of our data. So here each panel plots children by race so total represents all children and then white, black, Hispanic etc. represent different ethno racial groups for. The horizontal axis plots the number of children observed in the AFCARS in any given state in 1995. So dots to the right of each panel states that have larger population counts and states to the left of each panel have smaller population counts. And send the vertical axis plots the ratio of the count of children from the VCIS, the other source from that year to the AFCARS. And so if the two sources correspond perfectly that.will fall precisely on that dashed line at 1. So the ratio of the VCIS the AFCARS will be 1 whenever the counts are the same. And so if a dot falls above the line for it means that the VCIS enumerates more children than the AFCARS, so the dot falls below the line the VCIS enumerates fewer children than the AFCARS. And then the color of the dots just indicates two different kinds of measures so the blue dots represent measures of entrances of children into care in that year and then the red dots are point-in-time custody measures. And so you see a few different things. First of all you see that the dots toward the right are closer to the line and that's to be expected. What that says is in states where there are larger populations the discrepancies are smaller. So if are missing two or three kids in a state where we're enumerating thousands of kids that out a big difference. If were missing two or three kids in a small state with a small ethno-racial population then it's going to lead to a larger proportional difference. The second thing you see is that the measures for custody are generally more reliable than the measures for entrances. So the red dots to the line that the blue dots. That's also to be expected it's generally just easier to count the stock of children in care at any point than the number of children entering care over the course of a full year. The third thing you see is that's the dots are kind of evenly scattered above and below the line. So what that tells us is that the AFCARS and the NCANDS don't seem to be systematically biased in one direction or another. There's a lot of noise in any given state, the VCIS could overcount or undercount the population relative to the AFCARS. But when you average it out, the mean difference between the two sources is almost precisely 0. So this is to say we were making comparisons between say 1994 and 1996 in a particular state we might be concerned that any change might result from unreliability in the data across the two sources, but on average when we combine information from many states we think that there's not going to be a systematic bias upward or downward as we transition from the VCIS regime to the AFCARS regime. Okay and so once we've smushed all these data together we can create pretty long time series that illustrate really long term historical trends in children in substitute care. So here are the x-axis is historical time, the vertical axis is the rate of children in care so a population rate. What we've done here is taken a count of children in care and divided it by the number of children in that state in that year. I should say these are not age-adjusted rates. Age standardization would be something you would want to consider doing and it's something you could do with the AFCARS if you aggregated by age. It's something that would be harder to do but not impossible with the VCIS which reports information by age group. It's not something you could do with the Children's Bureau Statistical Series because we don't have any information about age for those children. But simply by plotting these trends by state over a 60 year period we can learn quite a bit about the dynamics of children in substitute care. For example we can see quite easily that consistently over a long period of time, the rates of children in substitute care in the South have been much lower than in other parts of the country. You can see trends in large influential states that mirror each other. So for example in California, in Illinois, in New York you saw large increases in the substitute care population in the 1990s that peaked around the turn-of-the-century and have markedly declined since then. In other states, Kansas, Montana, West Virginia, Indiana, you're seeing dramatic increases in the substitute care population that are as large or larger than any increases we've ever seen in observable history. I should say black dots here represent observed data so there are years in which we observe the counts of children. The red dots represent estimates so here what we've done is build a really simple interpolation model where we it's sort of a combination of linearly interpolating the trend in any given state but also borrowing information from neighboring states with non-missing data to make guesses about the value of children in care in any given year. You could develop any number of different procedures for making estimates of missing data. You could use a multiple imputation model, you could just use a smoothing estimator there are a number of general ways you could do it. The important point is that you do it because for example if you're interested in constructing a national rate and you didn't use estimates of state years that have missing data, those national estimates would be biased insofar as states were entering and leaving the sample in a non-random way. So for example here is a national trend that combines observed values of state-level substitute care populations but also includes in those totals estimates for state years with missing data. And so this sort of color band to illustrates how many states are contributing to the estimate each year and we can kind of see in the mid-1970s and the late 1990s that's when we're relying on fewer states to generate those estimates. We're using a larger amount of estimated data than observed data and because those points are local peaks we might be more suspicious about those estimates. If we used a statistical model to generate our estimates we could actually generate sort of confidence intervals around those estimates. But however you want to do it, if you are careful about thinking about missing data you can construct national trends and you can learn interesting things about the evolution of substitute care in the United States over more than a half-century. okay well of okay well I'll say more about what you can do with historical data next week. Today though I've just kind of shown you why you might want to think about doing historical analysis. If you decided you did, how you might think about leveraging the Archive's data products to do that and then what kinds of things you could start to learn just by describing the data. Next week we'll build a model of the relationship between a number of different factors: prison rates, poverty rates, and outcomes of interest like children entering into substitute care. We'll see what the relationship between those factors looks like and moreover, whether the relationship between those factors has changed over historical time. So hopefully if you found this presentation helpful for this week you'll join us next week. Thanks for attending thanks for listening and I'm really excited to take your questions. [Clayton Covington] alright thank you so much Alex for that really engaging presentation. I know that's really exciting for us to see you know the developing historical data holdings we have here at NDACAN. So I'll get us started with the questions. The first question we have says what were the specific differences in the ways that race and ethnicity were measured before and after 2000? [Alex Roehrkasse] great question! So this is a change that was general to a number of different federal data. The gist of it is that beginning around 2000 most federal data collection systems started allowing people to identify as having multiple racial identities. So before 2000, the set of options if you had asked questions about race for example, you would be forced to choose between white, black or African-American, American Indian, Asian or Pacific Islander, but there was no option to select multi racial categories. And so multiracial-identified people before 2000 are forced into a single race categories. After 2000, individuals are allowed to identify as having multiple racial identities. And so what that means is that, for example, counts of say white non-Hispanic children before 2000 include anyone who would identify as white and some other race. Whereas after 2000, generally speaking, people who identify as who are enumerated as white are people who are enumerated who identify as white only. And so want in particular to be careful about certain populations that are more likely to have multiple racial identities. But that's that's the gist of the problem. For that reason the SEER data for example can pose problems. There's a lot of demographers who have spent a lot of time trying to figure out how to harmonize racial data spanning the year 2000. Ironically because the Archive's data don't harmonize racial identity over that period, it's inadvisable to use harmonized racial data to construct your denominators. So whenever I'm doing historical analyses that include analyses of racial differences I use unharmonized racial data and use those to construct denominators for rates. And then you just need to be very careful when examining change that might happen over that threshold where race starts to become measured differently. In some states you to see some continuities that are almost certainly attributable to changes in the measurement of race. In other states for certain populations it's not a meaningful change in so it's not terribly concerning. It'll just depend on your specific research question. [Clayton Covington] alright thank you for that Alex. Our next question asks are the data available for free? And I can answer that very quickly by saying that all of the data holdings here at the national data archive for abuse and neglect are available for free. But we do have one more question that asks or actually a couple of more questions. The first one asks, which data sets will you be utilizing in your presentation for next week? [Alex Roehrkasse] yeah so next week I'll be talking about long-term trends in children in substitute care so I'll be mostly talking about the same three data sets that I used to construct these long-term timeseries, that is the Children's Bureau Statistical Series, the Voluntary Cooperative Information System, and no, I take that back! Next week since I'm going to be sort of talking about regression analyses I think it'll be confined to a combination of the VCIS and the AFCARS. So that'll be sort of analyses that span the early 1980s through today. And that allows us to do sort of can combine those data then with sort of predictors from a number of different sources. So I'll be talking about and how you might combine data, what kinds of sources might be helpful to combine with archive data. So yeah it'll focus on the VCIS and the AFCARS next week. [Clayton Covington] alright our next participant actually has two questions so I'll read them separately. The first question asks, based on one of your earlier slides is it reasonable to conclude then that we should be more careful in analysis and inference on flows in and out of care since there are greater discrepancies across the VCIS and AFCARS? [Alex Roehrkasse] the short answer is yes for two reasons at least. One is that as I illustrated the measures are less reliable both across sources and I think we should think within sources across states across years. Also there are just more missing data for the flow constructs. So more states report information about children in custody then do about children entering care. [Clayton Covington] alright. And the second question asks, do you know of any centralized documentation or database that collapses information across states and jurisdictions and time of what particular of what particular measures mean, for example how abuse and neglect is defined across states and counties, criteria for substantiation across states and counties? [Alex Roehrkasse] that's a great question. This is a major challenge to research in this area. I'm not aware of any single source compendium that would give directly comparable sort of information about constructs and sampling frames for different states. I would say the place to start is to look at the codebooks for archive data. The VCIS for example were published annually with quite a bit of metadata and documentation about which states were reporting what kinds of children. So but it simply takes a lot of elbow grease to go through the sources and see how the reporting works differently in any given state in any given year. It's our hope to combine the CBSS, the VCIS and collapsed AFCARS data into a single data product that would be available through the Archive. So that the kinds of analyses adjust showed you would be trivially easy to do for researchers who just wanted to download the aggregate synthetic historical data product. And of course that data product would have to come with quite a bit of documentation about comparability over time. But I don't I think if you go out beyond our resources to document state-specific changes over time in for example the measurement of the sort of procedures for substantiation or something like that. Just so it's a limited Lee helpful answer I apologize. Unfortunately I think it's just a major challenge in research in this area. [Clayton Covington] alright thank you for that Alex. We have another question that's a follow-up question. It's asking, do we have to get approval for getting AFCARS and VCIS data? [Alex Roehrkasse] yes you do. The AFCARS has a sort of well-established process for getting access now that's well-documented on the site. I'm actually not that familiar with it but Clayton you might be able to speak to it better than me. The VCIS isn't yet available through the Archive but we'll be developing procedures for accessing that in the coming year because the data is already aggregated it doesn't really pose any disclosure risk. So were hoping that we can make that data is easily accessible as possible. [Clayton Covington] yes and just to add, and feel free to chime in Andres because I know you administer a lot of the actual dissemination of data sets but there is a process of getting approved where you have to sign like a user's agreement. Also if your home institution you may need to go through IRB in order to access the data, but the approval process is pretty straightforward and my imagining is once the VCIS data is available on our website it wouldn't differ all that much from our current other administrative data sets in terms of obtaining access. [Andres Arroyo] yes this is Andres, archiving assistant, just want to say that yeah to get the AFCARS datasets you submit what's called a Terms of Use Agreement form and that's downloadable on our website, thank you. [Clayton Covington] thank you for that Andres. So there aren't any more questions as of now so if Alex if you could go to the next slide. So just as a reminder this is, oh there's a slight error in this slide but next week same time we'll on July 15, not the eighth, we'll have Alex again presenting using a specific research example using historical data, using the AFCARS and the VCIS as he mentioned earlier. But yeah I believe that's it so thank you everyone for your participation in the NDACAN Summer Training Webinar Series. It's been a pleasure to have you all and I think you for some really interesting questions and we hope you all will join us for the remainder of the series. [Alex Roehrkasse] I'll just say thanks to everyone again for listening and I know it's sometimes hard to come up with questions on the spot. If you have questions later on that you want answered, feel free to shoot me an email you can find my contact information on the NDACAN website. [Clayton Covington] alright, thank you again everyone, have a great afternoon. [Voiceover] The National Data Archive on Child Abuse and Neglect is a project of the Bronfenbrenner Center for Translational Research at Cornell University. Funding for NDACAN is provided by the Children's Bureau.