MCHB/EPI Miami Training — December 5 - 6, 2005

Group Discussion on the Use of SaTScan — Transcript

RUSSELL KIRBY: Yeah, this is from yesterday. Take a copy if you want. Okay. There's a couple of additional points that we wanted to discuss just briefly. One of them is this whole business about fields that might be in your data set that purport to tell you something about the geography.

So if you have — say you've got a birth certificate file and it has data on all the births in say in your state there's a field in your file for the county of residence. There's a field in your file for county of occurrence. There's a field on your file for the minor civil division, usually, in which the mother is a resident. There's a zip code field. There might actually be two zip code fields. There might be a potentially separate field for the zip code pertaining to the mother's residence. There might be a separate field for the mother's mailing address. If you're in a state that has a very backward vital statistics system they may have only one field which is a major source of problems in terms of geo coding the records.

Anyway, what I wanted to mention is what can potentially happen, and this particular set of slides that I have was given to me by a colleague from the Colorado Department of Health and environment. And they were doing a study a few years ago, the Rocky Mountain Arsenal. And the Rocky Mountain Arsenal is a place where the Army used to store a lot of nasty things that probably shouldn't have ever been manufactured. And so it turned out to be a toxic waste site. So they were doing an analysis of health effects in the vicinity of the Rocky Mountain arsenal. And so this map here shows kind of where the Rocky Mountain Arsenal is and I don't know exactly how they define the buffer area. But they define an area in proximity to it that was kind of the target area they were studying. And so superimposed on that map is the geography of the five digit zip codes.

So you can look at this and see that there's one zip code, the RMA, 8022 which actually captures the majority of this particular territory. But it also includes some other areas that are not part of the region. And then in order to capture the area, actually this one here you probably would want to include too, because it only has a little piece. And probably these two, also. But the point is that if you use just zip codes, you are going to end up needing to either lose some of the study area or include some areas that maybe aren't part of your region. Now, one of the things you could do if you were really crafty is link your geo coded data to the zip plus four and then you could potentially map it out more specifically.

Most people in public health don't go to that level of detail. And if you don't like zip plus four, just wait a year or two and they'll have zip plus 6, so you'll have even more spatial specificity in the postal data. This is a pattern that you get using zip codes. So now here we have the same thing with census tracts. The same area, basically, and you can see here that there's a lot more census tracts than there are zip codes. But they don't match up really all that well. In fact, there's this one core zip code that's there. But there are a whole bunch of others and there are many that cut across the region. If you use these two that are pretty large areas of the region, they're not going to match up very well either. So it probably isn't an ideal thing to do. So what they decided to do is why don't we geo code all the records and then we'll know exactly where they are. And we'll use that as a basis for classifying the records as in the target area or not.

Well, so they did. And what's really interesting about the analysis is that what they found when they did this analysis, because they had selected only the records that were indicated as being in the city, you know, that was that region. So what happens when you select just based on the city field on the data set, is all of these pound signs are indicating individuals who actually didn't live in the community even though their birth certificate, based on the geo code they didn't actually live in the community.

Now, I have seen this problem in virtually every analysis that I have done where I have been trying to geo code records that there is at least some level of mismatch between the spatial data that are reported in those specific fields and the actual locations that records geo code to. And in fact I know some people who work with these data enough that what they do actually is they create new fields for county of residence where they reclassify all the records based on the actual geo coded location that they got from the addresses, and they don't always come up with the same frequencies as you would otherwise.

Most vital — the people that run vital statistics agencies tend to want to rely on the directly reported information, but sometimes there's information in the records that's more specific that might be useful to look at.

So anybody encountered a similar kind of problem to this in any work that you've done?

UNKNOWN PERSON: Well, just where I live, I live in the city (inaudible) and a lot of people to (inaudible).

UNKNOWN PERSON: Can you turn on the lav?

Can everyone hear me? Okay. People who live in the bordering cities just tend to save Madison even though technically they don't live in Madison they're not in our analgesics so we spend a lot of time cleaning our birth certificate data up, for example.

RUSSELL KIRBY: It's also a problem in suburban fringe areas that are rapidly growing and sometimes jurisdictions change and people don't realize it. Like they might live in an unincorporated area and then it gets incorporated but they're so used to calling it the other thing. And you know we could have actually spent a whole day talking about issues around, you know, spatial classification of data. But there's a variety of issues that come up with this. And I guess the object lesson about this is to, you know, do take some time to study your data. But on the other hand, if you obsess over it, you won't ever get anything done because you'll just spend all your time worrying about these details. So at some point you've got to decide to fish or cut bait, decide that it's time to get on with the process.

So I just wanted to mention that. In terms of working with data sets, you know we've illustrated here most of the examples using vital statistics. I'd just like to ask the group, how many of you have done GIS analysis using other data sets and what those were and a little bit about your experiences and if anybody wants to offer any insights about that, I can pass this microphone around or —

So nobody else has used any other data?

UNKNOWN PERSON: Sure. Clinic sites and Medicaid data.

RUSSELL KIRBY: So Medicaid data. And what exactly did you do with the Medicaid data? You simply tried to —

UNKNOWN PERSON: We've done studies with family planning. We've done studies with EPSTD clients, eligible versus eligibles who received a service, dentals, a variety of them. We went through the EPSTD catalog.

RUSSELL KIRBY: Okay. Anybody else? Has anybody done any GIS analysis with WIC data? You've done WIC data. Okay.

UNKNOWN PERSON: Maybe it's easier to ask what you haven't done.

RUSSELL KIRBY: What haven't you done. She said I don't know but I'm working on it. Yeah. But —

UNKNOWN PERSON: We're going to be starting a project soon for children with special healthcare needs and the data that's related to children with special healthcare needs.

RUSSELL KIRBY: What exactly do you think you're doing with that?

UNKNOWN PERSON: Initially we'll start mapping where the services are and where we should be locating our clinics, but eventually we'll be doing some other kinds of things that have more to do with not just measuring where buildings are but whether or not we're actually getting the services out there.

RUSSELL KIRBY: I actually saw an announcement, dine probably did it because it's from North Carolina but I saw it presented a few years ago where they had looked at, they basically linked the birth defect data with children with special healthcare needs data and for specific regions of the state they looked at the proportion of the babies that had particular conditions like downs syndrome, like I think there might have been facial clefts, neural tube defects or spinal bifida and they looked to see how many children within the different regions were actually participating in the children with special healthcare needs program and on the basis of that they tried to target different areas of the state more heavily to increase awareness about the program and its services.

How about blood lead. Anybody work with GIS with blood lead screening?

UNKNOWN PERSON: I'll say a few words. We've done some work with lead screening levels and then trying to geo code the addresses off those records. And it's difficult. Address standardization is a big issue and then missing data. You know we've worked with WIC data and lead screening data and you know if you're coming up with 20% of your records with missing address data or that you can't geo code it down to a low enough geographic specificity to be useful, sometimes it's difficult to use.

RUSSELL KIRBY: Couple comments about that. The first one would be that using some of these more service oriented data sets statewide can potentially lead to issues, even yet there are still you know better data in urban areas than rural areas, for example, and we still you know I think that the coverage for geo coding are better than they were, but it's still a potential issue. Now, in a very urban state, you know, like Rhode Island or Connecticut that might not be as much of an issue but in a larger state it's a potential issue.

The other point I would bring up when you start working with these data to remember that the data fields that you're working with are not necessarily priority areas for the program. And in fact you might be the first person who has ever used those fields and so it's probably worth incorporating a data quality feedback loop and providing information back to the program on how — obviously they're not going to spend the resources to clean up the old data, which is they don't have the funds to do that. But you could potentially make recommendation as to how they could collect those data more effectively in the future so that over time you wind up with more usable data sets. And it's the same kind of principles that you would apply if you were talking to the people around your vital statistics agency about how to potentially improve those data. Lately I find it to be like talking to a brick wall but — how about birth defects. Anybody done any GIS analysis with birth defects data? . That's too bad because I was looking for some examples because the end of January dine and I will be leading a training seminar just like this for people who work with birth defects. So we'll have to make up some examples, I guess.

What about, has anybody done any GIS work with hospital discharge data?

UNKNOWN PERSON: It's usually restricted to zip code.

RUSSELL KIRBY: Zip code, right.

UNKNOWN PERSON: (Inaudible).

RUSSELL KIRBY: Yeah. Now, there are some creative ways that you can get around that, however. Depending again, these are the kind of things that people who work in the state health agency can probably do, our university types probably would never get the data sets for it. But if you are in, for example, say you want to map the prevalence of pre-eclampsia in your state. You could potentially link the hospital discharge data for maternity care with the birth certificate data, and because you've geo coded the birth certificates, you then end up with much more spatially-specific information than you would have just from the hospital discharge record. Again assuming that the birth certificate addresses are relatively clean and geo-codable. But there are a lot of creative things you could do. I was thinking in August I attended, I don't remember if LACOTA was there. I know Chris Denk. Charles Denk, that's right, he was there. And it was a conference on record linkage of serial pregnancies, successful pregnancy outcomes. And I think there actually would be some very interesting potential analyses that you could do geo coding those records. And among other things you could look at the potential impact of change in maternal residence across pregnancies and potentially get some kind of better measures of you know socioeconomic status and how it might be changing over time and look at those.

And I've never seen anybody look at those kinds of things. I've never seen anybody use GIS with those kinds of SID CHIP history data. There's lots of opportunities to come up with creative approaches to using GIS in the United States .

UNKNOWN PERSON: Just one question, and about two years ago we geo coded some like, part of (inaudible) cardiovascular disease data and then we used like a GWR, geographic weighted regression, to do some analysis. I don't know what your comments on this software. I have not seen —

RUSSELL KIRBY: You know, I actually haven't used the geographically weighted regression models myself. I've read a lot of papers about them, and I think what basically the problem that they're trying to address is the fact that when you aggregate your data into ecological units and run a regression analysis on those ecological units, you are basically throwing away a great deal of information in your analysis. And you're throwing away information in two different ways. One way you're throwing away information is by the mere act of aggregating the data, you're coming up with information that removes potentially information about the amount of variance within each of your estimates. So, for example, say you're looking at low birth weight as your outcome. The low birth weight rate as your outcome. And in one unit you have a rate of 7% for low birth weight but it's because there were seven low birth weight babies out of 100 births.

And another unit you have a 7% low birth weight rate but it's because there were 700 low birth weight babies out of 10,000 births. Those pieces of information have very different variance associated with them. And when you use a traditional ecological analysis, all you have is the simple fact that the low birth weight rate was 7%. So you throw away that kind of information.

What people have done to try to model that is to use Plasan or extra Plasan regression to capture that variance where they're modeling instead of the low birth weight rate they're modeling the number of low birth weight births per total events, you know, as the dependent variable. That helps some but it still doesn't take into account the fact that each of your ecological units actually exists in space and that it's adjacent to other particular locations. And the geographically weighted regression is an attempt to model the spatial pattern of these ecological units as well.

So it's definitely something to look at. And I believe there's a new textbook by (inaudible) that is definitely worth taking a look at. I didn't put it on the list of the bibliography. But that would probably be worth taking a look at. That's definitely an emerging technique.

In terms of spatial statistics, it's a very rapidly he involving field, and many of the methods that people use today are less than 15 years old and still evolving in terms of the techniques.

Okay. What I'd like to do since we're getting close to tend of our time we'll try a little experiment now to see how good everybody is.

And we're going to play a game. And the game we're going to be playing is called GIS and MCH epidemiology Jeopardy. It's a version of Jeopardy that we're going to play. The way we're going to work this, I'm just going to say — there's a few more people on this side of the room. That's okay. You guys are just extra good. So we're going to — did we decide what we're going to name the teams?

UNKNOWN PERSON: X and Y.

UNKNOWN PERSON: (Inaudible) and.

RUSSELL KIRBY: Which one do you guys want Geodar or SAT Scan? So those are our two teams in this competition. Now, I have made — okay. In terms of some of the rules. Don't forget that to answer, provide the answers they have to be in the form of a question. What is? Or who is? Or when was? Those kinds of things. Okay?

What's the prize? I actually tried to get —

UNKNOWN PERSON: I tried but I could not get professor.

RUSSELL KIRBY: I'll tell you what, as a prize you can have another ice cream bar, because there's a whole bunch of ice cream bars left. But I tried to get tootsie roll pops but I couldn't get them. I apologize for that. We'll play the game and we'll see how it goes.

So we have five categories. And we have a category for mapping methods. We have a category for GIS functions. GIS public health personalities, what you see is not what you get. Or we have candy bars. Until they run out we'll give candy bars for correct answers. Then we have a special category of facts about Florida . Is there anybody here who is from Florida ? You probably shouldn't answer — you probably will know some of these, because they're relatively obvious. And so why don't we have somebody from the Geodar team pick an item and we'll start the game. You have to answer in the form of a question. And yeah, you have to put up your hand and Diane is our spotter so she's going to identify who is the first person to put up their hand with each of —

UNKNOWN PERSON: Anyone on the team can answer.

RUSSELL KIRBY: Yeah, but you have to put your hand up to answer. Anybody on the team can answer. Lori, do you want to.

UNKNOWN PERSON: (Inaudible).

RUSSELL KIRBY: What you see is not what you get for 100. So this free software provides effective tools for exploratory spatial analysis. What is...now, we have over here.

UNKNOWN PERSON: (Inaudible).

RUSSELL KIRBY: What is SAT Scan? That was the wrong answer.

UNKNOWN PERSON: It's the opponents that get to answer automatically?

RUSSELL KIRBY: That's reasonable.

UNKNOWN PERSON: What is Geodar.

RUSSELL KIRBY: What is Geodar.

UNKNOWN PERSON: (Inaudible).

RUSSELL KIRBY: They did answer. Okay. And we have a bonus question anybody can answer this. This will known GIS scientist developed the Geodar software. Back here.

UNKNOWN PERSON: Who was Luke Anslon.

RUSSELL KIRBY: That is correct. So that means whoever answered that, okay you give us the next choice.

UNKNOWN PERSON: GIS public health personnel.

UNKNOWN PERSON: 100.

RUSSELL KIRBY: 100. Okay. So some have argued that this individual was an early practitioner of noncomputer assisted GIS. I don't need to read the whole thing.

UNKNOWN PERSON: John Snow.

RUSSELL KIRBY: Who was John Snow and we have a picture of John snow here. Pick another.

UNKNOWN PERSON: Facts about Florida .

RUSSELL KIRBY: Facts about Florida for 100. Okay. This beautiful blond rarely seen outside green houses in most American states is the Florida state floor.

UNKNOWN PERSON: Hibiscus.

RUSSELL KIRBY: Hibiscus, that was not the information I got. Anybody wants to take a stab at what is the state flower?

UNKNOWN PERSON: (Inaudible).

RUSSELL KIRBY: Somebody over here.

UNKNOWN PERSON: (Inaudible).

UNKNOWN PERSON: Orange blossom.

RUSSELL KIRBY: Orange blossom.

RUSSELL KIRBY: I have to make a confession. I'm relying on information I retrieved from the websites I did not fully verify all of the answers. But the orange blossom is what — it had a list of all the state flowers of all the states that's what it said. So we'll hope it was correct.

So you answered the orange blossom.

UNKNOWN PERSON: Facts for 200.

RUSSELL KIRBY: Facts for 200. Okay. The state of Florida is known by this nickname. What is:

UNKNOWN PERSON: What is the Sunshine State ?

RUSSELL KIRBY: Okay. So over on this side again.

UNKNOWN PERSON: (Inaudible).

RUSSELL KIRBY: GIS functions for 100. Okay. These spatial data formats developed by ESRI for use with its software.

UNKNOWN PERSON: (Inaudible).

RUSSELL KIRBY: Okay. Yes.

Okay. You know, I think you're going to end up doing GIS function for 200.

When I — for some reason the skip patterns on a couple of the questions weren't quite right. I thought I corrected them. But I guess not. GIS functions for 200. Okay. This GIS technique identifies areas within a certain distance of a point or map element, what is a —

UNKNOWN PERSON: Buffer.

RUSSELL KIRBY: What is a buffer. Yes. Okay. 200 is done also.

UNKNOWN PERSON: Mapping methods.

RUSSELL KIRBY: Mapping methods for 100. This map element provides the reader with information necessary to interpret the map.

UNKNOWN PERSON: (Inaudible).

RUSSELL KIRBY: Metadata dictionary. It's a little more basic than that. Over here.

UNKNOWN PERSON: What is a —

RUSSELL KIRBY: What is a (inaudible).

Okay. It was probably a photo finish on getting the hands up. On this side, a question.

UNKNOWN PERSON: Facts about Florida .

RUSSELL KIRBY: Facts about Florida . What happened here? Okay. Here we go. Okay. This smaller city in the northern part of Florida is the state capital.

UNKNOWN PERSON: What is Tallahassee .

RUSSELL KIRBY: Surprisingly that is correct. But we have a bonus question. How many states, we'll give 300 points for this. If you can tell me how many states have capitals with longer names than the capital of the state of Florida . Longer than Tallahassee . What states are those. This is a geography 201 question. You could have probably answered this in fifth grade really well. Anybody?

UNKNOWN PERSON: I'll take a guess. (Inaudible).

RUSSELL KIRBY: You could just say —

UNKNOWN PERSON: (Inaudible).

RUSSELL KIRBY: Zero. Well, that's not actually correct.

Anybody else want to take a stab at it? You want to try.

UNKNOWN PERSON: (Inaudible).

RUSSELL KIRBY: I can't hear you.

UNKNOWN PERSON: What is one —

RUSSELL KIRBY: Jefferson city Missouri . Okay. You know I think I missed Jefferson city .

But there actually are three others names that are longer. But Jefferson city . Yeah, Jefferson — so there's actually four. There's four that have longer names.

Anyway, that was facts about Florida . And you guys had it. So somebody on this side pick a category.

UNKNOWN PERSON: GIS functions for 200.

RUSSELL KIRBY: I think we did that one but I'll just make sure. We did that. And we'll go — so 300. In ArcView this tool must be used to create maps for publication or distribution purposes. What is a —

UNKNOWN PERSON: (Inaudible).

RUSSELL KIRBY: Layout. What is a layout. Right. Okay.

Over on this side pick another category.

UNKNOWN PERSON: Mapping methods for 200.

RUSSELL KIRBY: Mapping methods for 200. Okay. This map element often left off of maps informs the reader how far apart — spotter.

UNKNOWN PERSON: What is a scale.

RUSSELL KIRBY: What is a scale. Yeah. What is a scale.

Okay. So on this side somebody pick a category.

UNKNOWN PERSON: Florida for 400.

RUSSELL KIRBY: Florida for 400.

UNKNOWN PERSON: He went to the University of Miami .

RUSSELL KIRBY: That's right. But he probably was gone when this happened. So these events, an example of which brought state notoriety in November of 2000 are a hallmark of democracy but also sometimes reveal its flaws, what are —

UNKNOWN PERSON: (Inaudible).

RUSSELL KIRBY: Huh?

UNKNOWN PERSON: (Inaudible).

RUSSELL KIRBY: This is pretty easy.

UNKNOWN PERSON: (Inaudible).

RUSSELL KIRBY: Okay. Hanging (inaudible) you know, that isn't the answer I have. But those are probably good enough. I actually just — I was being much simpler and I was saying what are the Florida elections, but hanging chads are probably a better answer than that.

UNKNOWN PERSON: So does that count?

RUSSELL KIRBY: I guess so. Okay. So on this side pick a category.

UNKNOWN PERSON: Mapping methods for 300.

RUSSELL KIRBY: Mapping methods for 300. Okay. Okay. This type of map presents spatial data classified into quantitative categories displayed across administrative units. What is a — did you have anybody? Somebody on this side. Was it the girl in the back or —

UNKNOWN PERSON: (Inaudible).

UNKNOWN PERSON: He raised his hand first.

UNKNOWN PERSON: What is a corporate map.

RUSSELL KIRBY: What is a corporate map. Right.

Okay. Another category.

UNKNOWN PERSON: (Inaudible) personalities for (inaudible).

RUSSELL KIRBY: This one is a daily double. All right. Okay. So my sound is not turned up really loud, but it actually does have one. So this person founded ESRY, developed the original ArcInfo software and played a major role in the evolution of information science throughout the world. Who is — who is Jack Dangerman. And there we have him. All right.

So on this side over here. Pick a category. So U.S. functions for 400.

All right. So this methodology enables the user to visualize his or her spatial data in order to better understand its spatial distribution and spatial property.

UNKNOWN PERSON: (Inaudible).

RUSSELL KIRBY: That's too specific of an answer.

UNKNOWN PERSON: (Inaudible).

RUSSELL KIRBY: What is spatial analysis. Okay. Well, that's not — that's too broad.

(Laughter)
Yeah, over here.

UNKNOWN PERSON: (Inaudible).

RUSSELL KIRBY: Yeah. Exploratory spatial data analysis. ESDA. Okay. So on this side pick a category.

UNKNOWN PERSON: Maps don't lie.

RUSSELL KIRBY: Maps don't lie. We'll find out. In order to create corporate maps, data must be grouped, natural breaks, quintiles, mean and standard deviation, all examples, what is —

UNKNOWN PERSON: Class.

RUSSELL KIRBY: That's I think that's close enough. The answer I have here is what is classification. And all those are examples of how you can classify your data.

Okay. So on this side.

UNKNOWN PERSON: Mapping methods for 100.

RUSSELL KIRBY: That's techniques allow — what are projections. I think I wrote map projections. That's correct. Okay. So on this side.

UNKNOWN PERSON: The last Florida .

RUSSELL KIRBY: And you're going to have to be watching really closely because people will know the answer many almost before I put the question up. This person, an outstanding member of the MCH epidemiology community recently relocated to this state. Who is — you know the answer?

JOEL SEARS: Bill Sappenfield. I guess I've been doing this for too long.

RUSSELL KIRBY: It is. We have a picture of Bill Sappenfield here. We'll give both teams — since the Distance LEARNING people got the answer, we'll give it to both. But so do you want to pick a category? No. Okay. You have to pick a category now?

JOEL SEARS: (Inaudible).

RUSSELL KIRBY: Yeah.

We're getting down to it.

UNKNOWN PERSON: Florida .

RUSSELL KIRBY: We're done with Florida . So the ones in yellow are left.

UNKNOWN PERSON: (Inaudible) I'm a wild card. GIS public health personalities.

RUSSELL KIRBY: Okay. This person developed the SAT scan software for exploratory of clusters. Do you know the answer?

UNKNOWN PERSON: Martin Kohldorf.

RUSSELL KIRBY: I couldn't find a picture of Martin I looked all over the Internet. I couldn't find one. Sorry about that. I'll have to send him an e-mail asking him to send me one. This side pick a category.

UNKNOWN PERSON: WHIZ and wig.

RUSSELL KIRBY: Sat scan is an application designed to detect these areas. I think it was Kathy in the back there.

UNKNOWN PERSON: What are spatial clusters.

RUSSELL KIRBY: What are spatial clusters. Yes. Okay. So —

UNKNOWN PERSON: That was (inaudible).

RUSSELL KIRBY: Yeah. Okay. So what's the score right now?

UNKNOWN PERSON: It's too close to call.

RUSSELL KIRBY: Too close to call. We'll let him tally. But Kathy you get to pick another category.

UNKNOWN PERSON: Maps don't lie.

RUSSELL KIRBY: Maps don't lie for 400. Okay. These map elements detract from the readability of data graphics including maps. What is — I will point out we did not actually discuss this in the seminar, but some of you probably know this term. And as a hint, it is a term that was widely popularized by Edward Tufty in books like the Visual Display of Quantitative Information.

UNKNOWN PERSON: What is clip art.

RUSSELL KIRBY: No, that is not correct.

UNKNOWN PERSON: What is grids.

RUSSELL KIRBY: No. Although grids would be an example of this. Definitely. Definitely. But what's the generic term for it.

UNKNOWN PERSON: Junk.

RUSSELL KIRBY: That's half of it. Okay. The answer is what is (inaudible) junk. But, no, they don't get that. However, since he tried to answer we'll let him pick a category, okay.

UNKNOWN PERSON: Personalities.

RUSSELL KIRBY: Personalities for 400. Okay. This person creator of public health GIS news and information newsletter has greatly popularized the use of GIS —

UNKNOWN PERSON: You must be a subscriber.

RUSSELL KIRBY: Who is —

UNKNOWN PERSON: Charles (inaudible).

RUSSELL KIRBY: Kroner, right.

UNKNOWN PERSON: Can I report?

RUSSELL KIRBY: Yes, please.

UNKNOWN PERSON: (Inaudible).

RUSSELL KIRBY: That's what I expected and that's what's supposed to happen. That's good. It's a form of Zen sports, where the purpose really is to come out even. But we'll let this side pick one of these 500s. Right now they're tied. I don't actually have — if you guys split these last four, two and two, I don't have a tie-breaker question. So you'll really have to be on your toes to get these. But this side can pick a category. Mapping methods for 500. So spatial data must be referenced according to one of these systems in order to be displayed using a GIS.

UNKNOWN PERSON: What are coordinates.

RUSSELL KIRBY: I don't know about these guys here. Maybe I didn't make these hard enough. Do you think they're hard enough? Maybe, we'll see. Some of these last ones are going to be easy. But pick —

UNKNOWN PERSON: Personalities.

RUSSELL KIRBY: Okay. So this person the leader of the ostensibly Omniotic band is well known for a series of hits most recent —

UNKNOWN PERSON: (Inaudible).

RUSSELL KIRBY: That's right. That's the answer. In case you're interested, this is actually off of my web page at the university, and you can go to the URLs listed up you can put Russell Kirby Top 10 and it will come up. But I've got about 20 and I'll have to have my colleagues over the last time today a little bit but it will posted on there as well. We have two more. Pick GIS or WHIZ and Wig. A GIS is more than software and data. Without these people who will do the work. What are?

UNKNOWN PERSON: (Inaudible).

RUSSELL KIRBY: What are epidemiologists? No. No, that's not quite right. What are — yeah, anybody want to try an answer?

UNKNOWN PERSON: Geographer.

RUSSELL KIRBY: That's not quite right either.

UNKNOWN PERSON: Program analysts?

UNKNOWN PERSON: Digitizers.

RUSSELL KIRBY: Have you heard it yet?

UNKNOWN PERSON: Spatial.

RUSSELL KIRBY: Spatial analyst is close. That's close. But actually the answer is GIS professionals.

UNKNOWN PERSON: Oh.

RUSSELL KIRBY: And it really is important you know when you're building a complex GIS to think about having people with specific training in GIS. So we've got one more question. And it is GIS functions for 500. So I'm going to click on this. And you can all put up your hands because this is the end. Okay. Okay. So this process is an essential step in converting address data. What is geo coding? And that is the game. How did we come out.

UNKNOWN PERSON: Geodar team has 500 points.