[Voiceover] National Data Archive on Child Abuse and Neglect. [Erin McCauley] Hello everyone, and welcome to the second to last presentation for the Summer of NYTD series! This summer training series is brought to you by the National Data Archive on Child Abuse and Neglect in the Bronfenbrenner Center for Translational Research at Cornell University. As you can see in our schedule, this is the first of the research presentations which we’ll be ending out series on. I’m going to be discussing an on-going study I’m doing using the NYTD outcomes for Cohort 1. Next week we’ll finish off our presentation by an accomplished professor who has used our data extensively. She’ll be doing a conference style presentation on a completed research project as well as discussing her experiences publishing off of our data. But as I said today I’ll be presenting on an in-progress research study. I’m a graduate researcher with the BCTR working on the National Data Archive for Child Abuse and Neglect where I do research with the NYTD dataset and then I also do kind of outreach with the data set such as coordinating and hosting this series with Michael Dineen. But today we’ll be going over my progress so far on an ongoing study, I’ll do a brief conference style mini-presentation of what I’ve found so far, and then review the next steps I plan to take in the study, and then we’ll finish, as always, with a  Q&A session with both myself and Michael Dineen, the data manager for the NYTD dataset, we’ll both be available to answer questions.  So I started this project knowing I would use the NYTD dataset and hoping to create a research question that would leverage the unique linking possibilities of NYTD. Beyond that I was relatively open to where the project would take me. So I started with a general literature search reading studies related to the transition to adulthood for youth aging out of foster care. Then I took themes from the literature to generate ideas, picked a question and did a descriptive analysis, I have completed an initial analysis, and then I will refine my analysis with the direction of my advisor and any feedback you guys have for me but my next steps will focus on dealing with missing data, exploring the role of mediating variables, and perhaps some differentiation but that will kind of depend on sample size. So to jump in to conduct my literature search I started with CanDL, which is the publicly accessible Zotero database used by the archive which includes all of the research studies that use our data. You may remember this resources from Session II, where we discussed the data user supports available through the archive. After that I explored sociological literature related to aging out, then looked at psychology and human development databases, and last I looked at some interdisciplinary databases such as policy, public health, and social work. I used the following key words—foster care, well-being, transition into adulthood, aged out, and institutionalization. From these key words I found articles that related to my topic, and then I also searched through their references to find other related articles, and used the feature to see what articles had cited the article I was looking at to find additional articles. That kind of last step of looking at who they cited and who cited them got me a considerable number of the articles. And this yielded a rather large body of literature on aging out of foster care system and the transition to adulthood. From this body of literature I distilled several key themes, three of which I thought about seriously and therefore included in my presentation. First, disability was a key theme that emerged across the research. Many of the research studies I examined ignored disability, while others brought attention to the lack of information about how processes and outcomes may differ between those with and without disabilities. There were also a number of articles that addressed the high prevalence of disability among foster care youth. Another key theme was questions related to if local patterns hold up in a national sample. There were several clusters of robust research that were using administrative data from local areas—two common areas were the state of Oregon and a cluster of mid-western states. Leveraging the national sample of NYTD I could address if these patterns held up outside of these research regions. Last, I found a considerable number of the larger and more national studies used a cross-sectional research design. Another potential area of research could be examining of these cross-sectional results in a longitudinal data setting. After fleshing these themes out a bit, I chose to pursue the theme of disability. Thinking about how to leverage a large longitudinal dataset with a higher national representation, I developed questions that seemed to be missing from the existing literature which I found particularly interesting. Eventually I chose a research question that I felt leveraged the unique linking capabilities of NYTD. I chose, “what is the role of demographics, foster care experiences, and child protective services history in the relationship between disability and incarceration, homelessness, childbearing, connection to an adult, and substance use?”. I chose these outcome variables because of the importance of these outcomes for long term health, economic success, and wellbeing. And this was in the literature both for youth aging out of foster care for youth who just are becoming adults who had foster care experiences, and then just generally for adults as they kind of go through that 18 and up process these kind of stood out as important indicators. So next, with help from Michael Dineen shout out in the Archive he gave us our expert presentations and the linking presentation, we linked the NYTD, AFCARS, and NCANDS data and then formatted the data. As a reminder the AFCARS data captures youth experiences in foster care and the NCANDS data reflected the youth child protective histories. So using this data, which I’ll talk a little more about later, I ran a descriptive analysis. I was curious to see if the averages for a variety of potential control variables, a perhaps service receipt variables, and outcomes were different between youth with and without disabilities. A little over 40% of the sample reported having a disability, and that’s a fairly large percentage, which gave me pretty good-sized groups so I ran an orthogonality test to see if there were differences by the means. I’m not going to bore you with all of the tables but I will give you an example table and then point out the areas across the data that there were no differences which is surprising. So on the left I have an example of an orthogonality table using the NYTD outcomes. As you can see there are four columns in an orthogonality table. The variables the mean for youth without disabilities, the mean for youth with disabilities, and the P value of the orthogonality test. And so I was looking at the p-value on the right hand side to see if the means are different by disability status. So for the NYTD data which we can see on the left all but one variable had a p-value of less than 0.05, meaning that most all the variables were statistically significant. The one variable where the mean was not significantly different between youth with and without disabilities is public housing assistance. At first this kind of surprised me but then I thought back to what Tammy and Telisa said in presentation 1 about how and why the data is collected. And so I thought perhaps a lot of youth who aged out were receiving housing assistance with the independent living program, so perhaps for both youth with and without disabilities there were a high proportion that received public housing assistance. And this is a process that I really like to do early in a study it helps me explore the data sets know it and then think through why we might see differences in areas and no differences in areas. And then I looked at the outcomes with NYTD and AFCARS. I’m not going to show you that table but there were no areas that were not statistically significant by disability status. So looking at all the variables I had from AFCARS there was a difference by disability status. I saw that youth with disabilities had a higher numbers of placements, higher average number of days in foster care, higher average number of removals, a younger age of admittance to foster care, and there is a difference in the rural/urban scale. And then I also looked at reason for removal for the foster care data and found that children with disabilities had a higher percentage of removals due to sexual abuse, child alcohol abuse, issues related to the child’s disability, child behavioral problems, and child abandonment. But on the right hand I kind of summarize the variables with my kind of final data set that links all three that were not statistically significant between those with and without disabilities. We can see it’s public housing assistance which I just talked about, proportion of youth with CPS abuse that includes physical or sexual abuse, and the number of victimizations from the reports for CPS abuse. But overall I found that there was a difference by disability status for many of these including my outcome variables so I kind of thought there was clearly like a “there there”, so I kept going and ran some initial analyses which I’m going to present here today. So I’m going to present these kind of like a mini-conference presentation style so ignore my dorkiness in the beginning. Hi everyone, my name is Erin McCauley and I’m a graduate researcher with the National Data Archive on Child Abuse and Neglect. And I’m going to talk to you all today about a study I did this is kind of the early stages examining how disability shapes incarceration, homelessness, connection with an adult, substance abuse, and child bearing during the transition to adulthood for youth who are likely to age out of foster care. Psychologists and human developmentalists have long hailed the transition to adulthood as a pivotal function for the long-term success of youth. This is particularly true for youth who are at high-risk, such as foster care youth. Moreover, this transition can be especially difficult for youth who age out of foster care. The percentage of youth who leave the foster care by aging out, which is also sometimes called through emancipation, has been rising steadily. In 2010 around 11% of foster care youth aged-out. And these youth are at higher risk, and this transition to adulthood after leaving foster care is an important transition that has long-term impacts for their success and well-being. At the same time, youth who have aged out of foster care have disproportionately difficult backgrounds, even compared to other foster care youth. They are more likely to have experienced more transitions and greater placement instability, and are less likely to have family or social relationships they can rely on during difficulty. Previous research has found that youth who age out of foster care have earlier pregnancies, struggle with homelessness and housing, and are in contact with the criminal justice system, and report higher proportions with mental health problems. So I looked specifically at two different studies so we’ll go through those briefly. This is a study published in AJPH in 2013 that found that youth who age out of foster care are at particularly high risk of becoming homeless. Around 31-46% experienced homelessness at least once before turning 26 years old. They found that running away while in foster care, greater placement instability, being male, having a history of physical abuse, engaging in delinquent behaviors, and mental health disorder symptoms were all associated with an increase in the relative risk of homelessness. 37% of the youth they interviewed for that were in their sample experienced 1 or more indicators of a difficult transition to adulthood. And so you can see they kind of talk about processes like areas where people experienced difficulty and then also kind of more objective outcomes so while many people struggled with things like having enough money that maybe perhaps like a different percentage struggled with employment. So that’s kind of how they organized it. And so pulling from these different themes we can see a lot of the outcomes that I looked at are here. So we have homelessness, incarceration. And then I looked at another study. This was qualitative study on youth perspectives on aging out foster care and transitioning into adulthood. Youth identified the following areas as having particular importance in managing this transition. First, they discussed self-determination. So youth talked about how previous experiences of autonomy were really important. They also talked about how coordination and collaboration of services was important, and how reaching out for help was really difficult during this transition and if they were turned away for asking in the wrong place repeatedly they stopped asking. So having one place to go to to coordinate services across areas they found extremely helpful. They also talked about the importance of relationships—many didn’t have family, but connection to an adult or someone on their side was extremely important. And for those that had family connections, some discussed how having family was a support when trying to go through that transition to adulthood, whereas others discussed how families kind of acted as a burden instead—especially when family members were trying to rely on the youth for financial support during this transition, or if their family were involved in drugs or other criminal activities. Additionally, they discussed the importance of their foster care experiences for determining their success. Youth who experienced higher levels of trauma in the foster care experiences talked about how that made their transition more difficult. But also youth talked a lot about how normalizing their experiences with peer relationships with other people who had been through something similar was important. And then last, youth talked a lot about how having a disability complicated their transition, and that accessing care was really difficult during this transition especially if they had previously been given access to care kind of through being in the foster care system. And so I really focused it on that last area when kind of trying to draw upon existing literature for themes. So across research studies, a couple of areas came up repeatedly in examining the transition to adulthood—social support, mental health and substance use, child bearing, incarceration, homelessness or living arrangement, finances or employment, and public assistance receipt. In reflecting on these topics and what was available in my data, I chose a few to explore so I chose to concentrate on incarceration, homelessness, child bearing, connection with an adult, and then substance abuse. And those are all outcomes that are in the NYTD which we talked about during session 2. And so this study seeks to answer the following objectives. Is having a disability associated with the probability of experiencing incarceration, homelessness, substance abuse, connection with an adult, or childbearing among youth who are likely to age out of foster care. And then additionally, I’m interested in if this association persists when controlling for youths’ foster care experiences, and child protective services history’s. And then last, if this association persists when examining the probability of experiencing these outcomes after leaving the foster care system specifically. So for those of you who were here on session to we talked a lot about how the coding of the NYTD outcomes changed just between waves one and the later waves and so I’m most interested in the time period following the transition into adulthood. But I did also want to see if there’s a big difference in in looking at the cumulative probability so the probability that someone has experienced at any point in their lifetime .So to answer these questions I used linear probability models and the NYTD data or the National Youth in Transition Database from the National Archive on Child Abuse and Neglect. And with the assistance of Michael Dineen I linked the NYTD data set with AFCARS and NCANDS and I examined three concepts—disability status, foster care experiences, and child protective histories. I looked at five outcomes which I’ve repeated many times I’m not going to repeat them again but they are right there. And I examined if the youth reported having these experiences in their life, or after aging out. So if you attended session two, we discussed these differences but for those of you who didn’t, in the first wave of the NYTD outcomes they asked the youth if they ever experienced incarceration, homelessness, substance abuse, or connection to adult and childbearing and that wave is collected when the youth are age 17 so before actually age out. It’s kind of like a baseline and then the next two waves are after they’ve transitioned out and so then it’s looking at kind of between 17 and the older ages 19 and 21 and we want to see if these youth have experienced these outcomes and so I did look at both if we are looking at all three waves combined so if they’ve ever experienced it before aging out or after and then if they’ve experienced it just after aging out. So I’m going to talk a little bit more about the sample creation. So Michael Dineen, who gave the presentation last week about linking NYTD with AFCARS and NCANDS linked the data for me. So using the NYTD file, he linked participants back to their AFCARS and NCANDS histories. So basically he linked up the youth who aged out of foster care to their foster care experiences and then also their child protective histories. The steps we took to link the database, our as follows. First, Michael linked the NYTD with the foster care file. Then, since the data format of the child file is a little different, I created a list of variables I was interested in. A child may appear multiple times in the child file and therefore have multiple reports. So I want to kind of create summary variables to reflect their experiences. For example, using information from the existing literature, I thought it would be important to know if the child’s abuser was ever a parent. I was also curious if a child had ever experienced a certain type of abuse, and then I was interested in cumulative variables which may proxy a level of contact such as the number of child protective services reports. So I worked with Michael to create a list of summary variables that would be possible, and and then Michael created a table with those summary variables and then merged this table with the data that he had already merged between NYTD and AFCARS. So this was a pretty brief summary of how we came up with the sample, but if you’re interested more in merging and you missed last week’s session I recommend watching the video. So we recorded last week’s session just as we record all the sessions to turn them into videos which we’ll release towards the end of the month. And so Michael will really go through more in-depth what he did but we also already have a video that is on the webpage which if you want instructions if you were here for session 2 we also looked at the link to that but Michael will also be here at the end if you have questions. But for now we’re going to move on. So looking at the demographics table I have the means for the outcome variables by disability status. So we can see that those with disabilities had higher percentages experience homelessness and incarceration, lower percentages report pregnancy or substance use, and equal percentages reporting a connection to an adult. We can also see that the racial distribution is relatively similar, however there are slightly lower proportions of those with disabilities who identify as Black or other and higher proportions being Hispanic or Native. When we look at the foster care experiences we can see that those with disabilities had a higher average number of removals and placements, as well as spending more days in foster care on average. Then when we look at removal reasons, we can see that there are differences. For example for those without disabilities a higher proportion of people had neglect as the reason, but a higher proportion of those with disabilities had child behavior problems than those without. And then last we have the summary information from the child file. Those with disabilities had a slightly higher average number of victimizations, and higher average number of reports, and a lower proportion experienced abuse by a parent. This is some examples. So now we’re going to look at the initial results. Iran a bunch of different models kind of see what’s going on here. And so it is a little confusing because there is going to be a lot on the page so I’m going to go through a blank version of the tables we’re going to look at. First, just going to be a column where we’re looking at disability as a predictor. And so first I run a bivariate analysis where having a disability predicts the outcome in model 1, and then I added in controls for demographics in model 2, and I add in controls for demographics and foster care experiences in model 3, and then last controlling for demographics, foster care experiences, and child protective service histories in model 4. This allows me to see how the relationship between disability and their outcome varies when we’re able to control for different institutional histories. And so basically what I’m doing is where all the little betas are that’s going to be the coefficient for disability out of the different regression models I ran. And then below I just tell you what I’m controlling or not controlling for. But beyond just using disability is a predictor one area of research that I found in public health that was looking at disability in other contexts talks a lot about how types of different disabilities have different mechanisms. And so it might be important to differentiate by disability type. So I looked at emotional and mental disabilities next. And so basically I’m just comparing those with emotional or mental related disabilities to those without know disabilities using the same four models. And then I reran these analyses. I looked at physical or sensory related disabilities as the predictor. And then I did this for each outcome, so childbearing, connection to adult, experiencing homelessness, incarceration, and substance use. At this point I’ll present the models for you. So first we’re going to look at childbearing but I will go through one at a time so don’t get overwhelmed. We have any disability, and so we can see that across the models this relationship is significant. So having a disability is associated with with a decreased probability of childbearing. In model one we can see that this is about a 4% difference where folks with disabilities were 4% less likely and then as we introduced demographics so model 2, we can see that this association becomes smaller so it’s about a 3% reduction and then it comes back up as we add in controls for foster care experiences and child protective service histories ending not quite as high as the bivariate relationship but certainly higher than just controlling for demographics. And then the top is after aging out which is my main area of interest but I also ran the same analyses for looking at the probability of ever having a child across the bottom and we can see it’s a fairly similar trend, the numbers a slightly different but the trend is the same. And then I looked at emotional or mental and we can see that this also has a significant association with a decreased probability of having a child. The relationship is the highest in model one. It goes down when we consider demographics and then comes back up again as we consider foster care experiences and child protective histories so it’s extremely similar to the overall indicator for disability. Then I ran it again for physical or sensory where we do not see a relationship either after aging out or ever. So then I looked at connection to an adult and I did not find any association when looking by any disability so these associations were not significant across any model looking at ever or aging out. Looking at emotional or mental, I found the same: non-significant. And then when I was looking at physical or sensory in the main area of interest after aging out there’s also not a significant association however it was significant when looking at ever. So I found this quite surprising the kind of existing literature based on that I hypothesized that there would be an association here so that was quite surprising. Next I looked at experiencing homelessness so we can see for any disability that there’s not an association, however for emotional/mental related disabilities there is an association where having a disability is associated with about a 4% increase in the probability of homelessness in the bivariate model and that stays the same when I include demographics however once we include or control for foster care experiences and child protective service histories this relationship is accounted for by those differences. And then last we see a very similar thing with physical or sensory related disabilities where we have a negative association in the beginning and then it’s ameliorated as we move through so that association goes away when we consider the foster care and child protective histories. And so it’s interesting those are in different directions. Next we look at incarceration, for the bivariate there is an association where having a disability is associated with an increase in the probability of incarceration, but this did not hold up when I included demographics or foster care or child protective histories so that the relationship was not particularly strong. When we look at having an emotional or mental related disability we have more of a persistent effect so it stays when we include demographics as increased probability however it goes away when we include or control for foster care experiences or child protective services history which I also thought was quite surprising. And then for physical or sensory related we have a decrease in the probability of experiencing incarceration which is persistent across all the models meaning that this is a really robust finding. So even when we control for demographics, foster care experiences and child protective services histories having a physical or sensory related disability is associated with about a 10% to 11% decrease in the probability of experiencing incarceration. We can see that having a disability is associated with an increase in the probability of substance abuse but that it does not persist when looking at demographics, foster care, or child protective service history however for emotional or mental related disabilities there was a strong association that persist across the models. So you can see that having an emotional or mental related disability is associated with an approximately 2% increase in the probability of reporting substance abuse during the transition into adulthood. But for physical or sensory there is no association when looking at after aging out and there’s a decreased association when looking at ever. To summarize the results across all the many models and outcomes, thank you for being patient with me while we worked through them, I know that there’s a lot, that’s typically how I do the early stages of an analysis, but I’m going to go through some quick points so individuals with disabilities and emotional or mental related disabilities were less likely to have children. Individuals with physical or sensory related disabilities were less likely to be connected to an adult. The increase in the probability of homelessness associated with having an emotional or mental related disability and the decrease associated with having a physical or sensory related disability were rendered insignificant when foster care history was introduced. Increases in the probability of incarceration for those with disabilities and emotional or mental related disabilities was accounted for when we included our controls. However folks with physical or sensory related disabilities were less likely to be incarcerated than those without disabilities across all of the models. This kind of surprised me and so I was thinking about this it might be because this group is at a generally high risk, because we’re comparing youth who age out of the foster care system regardless of if they have a disability or they do not. I thought that was quite interesting. And then individuals with emotional or mental related disabilities were more likely to experience substance abuse issues than those without emotional or mental related disabilities are no disabilities at all, whereas individuals with sensory or physical related disabilities were either less likely or there was no differences across the outcomes. So I thought that was quite interesting. So in thinking about what this all means a few things jump out to me. First, there wasn’t a huge difference in examining the lifetime prevalence and the post-aging out prevalence. There were a couple areas where we saw significant relationships for the lifetime but not for the post-aging out but in general they were quite similar. However there were large differences when differentiated by disability type. In some cases there was even a sign flip. So the relationships were going in the opposite direction and this indicates that there may be different mechanisms at play. Individuals with emotional or mental related disabilities were at particularly high risk of substance abuse. Individuals with physical or sensory related disabilities were either less likely to experience outcomes or there was not significant difference. And last, I was expecting to see an association between having a disability and the probability of experiencing homelessness and incarceration among youth who age out. But moving forward want to think more about why I’m not seeing this association in there. And then last, there were almost no effects at all when examining connection to an adult and so I thought this was quite surprising and another thing to look into. But my biggest take away is that looking at disability broadly may obscure effects especially if they are in the opposite direction and so that was something that maybe I should’ve expected to see but surprised me a little bit see the opposite directions. And so when I think about my next steps I’m definitely interested in doing a few things as I move forward. I want to deal with missingness and weighting. Currently I just kind of did a listwise deletion and this may cause a downward bias. I’m definitely going to deal with that using some of the tricks we picked up earlier in the series. I also want to reexamine the models perhaps using a logistic regression just to see if there are substantive differences. I used the linear probability models so that I could compare the coefficient but I want to make sure that the individual outcomes between the different modeling strategies are substantively similar. I’m also interested in interaction effects. So I have some hypotheses about where there might be moderators, but I also want to go back in to the literature and make sure I’m not missing anything and make sure that I’m grounding my decisions what moderators to check in the literature. And last, I really would like to differentiate further if possible. So in a lot of the literature looking at disability the intersection of race and disability is kind of a fruitful differentiation to make. Because often they are quite different experiences for folks across different racial or ethnic groups in regard to how the outcomes vary by disability. I’m also interested in looking at a more granular disability type so I created two broad categories looking at emotional or mental disabilities and physical or sensory disabilities, but I could break those into four categories and just see what happens. It’ll kind of depend on the sample sizes I end up with however I think it is worth looking into since there was such a difference by those two categories. But that’s kind of what I’ve done so far, what I’m thinking and I would love to hear your guys feedback and then as always Michael and I are both here to answer questions. And just thanks everyone for bearing bearing with me through that. So Jamieson wrote “How are you planning on dealing with missing data exactly?” Multiple imputation. Yeah so that will definitely be my first plan of action is to go through and see first to look at that some of the tests that Frank put us through just to see what level of missingness is kind of random are not would be helpful. And then yes I plan on using multiple imputation. I haven’t started making an actual plan yet but that’s certainly my first plan of attack. Okay we have another question from Valeria: “great presentation and an interesting analysis. Since disability can be time variant at what age did you describe disability, the most recent records or cumulative? How far back into AFCARS and NCANDS did you track your NYTD population?” Fantastic question so yes experiences based on having a disability can be treated as time variant as well as having a disability so for my analysis I took the indicator of a disability from when the youth were in foster care. And so that I’m not sure exactly what point in time. Michael you may be able to answer that because he’s the one who pulled those indicators for me but it’s before the time period where we are looking at aging out. And then how far back into AFCARS and NCANDS did we track the NYTD population? And so it really depended on when the youth had their first contacts. So for some of the youth their first contact was very early in their lives, we picked them up at that point but for youth who perhaps their kind of first appearance in the data was later in their life then we picked them up later in their life. But for the cumulative ones we did consider like it starts from the very first moment of contact. Hopefully that answered your question Michael do you have anything to add in there? [Michael Dineen] different states have different years that you can go back to so it kind of depends on the state and I’d have to look at my code to see if I fixed it at some particular year or not. I don’t remember offhand. [Erin McCauley] yeah I don’t remember either but I do know that it was before the aging out period where we’re looking at the outcomes. [Michael Dineen] yeah and your one thing that I thought might be a little confusing on one of your slides it said sexual abuse and I’m not sure whether how people interpreted that but I thought it could be interpreted two different ways one is have they experienced sexual abuse after aging out, and the other was did they ever experienced sexual abuse in the child file portion and it’s that latter: did they experienced sexual abuse in their earlier at earlier in their life not after they aged out because there is no way of knowing whether they experienced sexual abuse after they aged out. [Erin McCauley] yes thank you Michael so that was that was under kind of the reasons area if that makes sense that was a control. Thank you Michael I’ll make sure to clarify that in the future. [Michael Dineen] I also while I’ve got my while I’m unmuted I wanted to make a suggestion to you about your follow-up study. Would be that since you had so many individual tests and with a .05 probability cutoff then with so many tests the likelihood of getting you know one in 20 that are going to be significant by chance is high right. [Erin McCauley] yes [Michael Dineen] and so what might really help is if you did the same analysis with the 2014 cohort which is a completely different group of people and see if the if those things that were significant remain significant with a different population. I think that would be a pretty solid, it’s like doing the same study on two different groups. [Erin McCauley] yes yeah thank you Michael that’s fantastic suggestion especially I mean right now we wouldn’t be able to go all the way through aging out because we don’t have all the outcomes but we would have enough and we could compare the lifetime prevalence. And just to clarify like the next steps won’t be a follow-up study it’ll just be me continuing this one this is very early analysis that I’m sharing with everyone so thanks for being so kind. So we have more questions: from Chris “wonder if it might be worth also considering differences by gender”. Excellent excellent idea, don’t know how that didn’t occur to me. But particularly when we when we look at the outcomes those tend to vary by gender by themselves but I’ll also for folks with disabilities that tends to be a lot of gender differentiation in the way that people respond to different behaviors for different genders so that’s an excellent idea thank you Chris. From Hughes sorry if I’m missing people’s names up I’m doing my best: “how did you address duplicate cases?” I’m assuming that you’re talking about the services file which I actually didn’t include in my final analysis. So if that is what you’re talking about then that is the answer but if it’s not then put another question in and I will address it again. Gloria: “Were there any variables on whether a child has an IEP in school and their classification, school interventions and how that might impact outcomes?” I didn’t, I don’t believe that we have any of those. If you want to see everything that I could possibly have worked with it’s in the codebooks for the three data sets but I think that might influence outcomes quite a bit. I do know that from reading the qualitative research in this area that one of the issues that youth in foster care experience who have disabilities is that they have more frequent placement transitions and so lots of times the schools didn’t even set up an IEP and they because they didn’t know that the youth had a disability while they were with them because the bureaucracy like by the time the parents found out that the youth had a disability that required an IEP they had already transitioned to a new school. So that’s definitely a big issue in this area and so finding out that information would certainly enrich the outcomes but as I’m aware I don’t think that we have them. But yeah school interactions would be a particularly fruitful area to move forward with these especially if we know that people’s experiences in foster care affect these outcomes that if we could capture them in schools and do interventions in kind of that same age period that would be extremely helpful. Hughes: “I believe you stated you used NYTD outcome. I noticed in the NYTD service file they’re reported twice” okay so there’s the NYTD outcomes and there’s the NYTD service file. I only used the NYTD outcomes I didn’t look at the services that people experienced but that might be a good idea and there might actually also be information about if they have if they received if they received special education services, that would be in that file so perhaps merging in the services file would be helpful. And speaking to both Hughes’ and Gloria’s questions, so I was totally wrong there might actually be information on that I don’t think that has that level of information but differently has if they received special education services. And then Jamieson said “Will you use imputation or any other methods accounting for missing data to produce descriptive statistics? Perhaps statistics describing the proportion of youth who have disabilities and are incarcerated vs. not.” I don’t believe so I generally I don’t think you’re supposed to use imputation on outcome variables so I wouldn’t apply it to the outcome variables. And with disability being the number one predictor I think my first round would just be to impute the other control variables but not the predictor and the outcome. But I mean you know this is early so I don’t know where it will take me but for now I plan to impute only the control variables in my three control variable categories. Alright I don’t see any more questions coming up. Thank you everyone for participating in this. This has been really wonderful. I was a little nervous to share such early results but I just wanted to show you guys what it’s been like for me just kind of working through this data and kind of dealing with the challenges especially with so many data sets in mind like I think the probably number one thing I’m going to do next will be did look at the services file and look at if we can kind of as some information about if that special education services while in school I think that would be extremely helpful so thanks thanks for that lead. But we hope that you guys can join us next week. Svetlana Shpiegel is going to be doing our last presentation. She uses our data a ton. Were very very lucky to have her giving presentation. She’s going to be talking about a published study that she’s done with little bit of detail and then she’s also just going to be talking about her experiences in publishing with our data and kind of the areas where she finds she has to explain the data to people. So for those who plan on doing research projects with this data I think it will be a very fruitful discussion. And it will be our very last session so we hope you close it out with us. Thank you everyone for participating, this has been quite a session. [Michael Dineen] Thanks everyone. [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.