Case1

Summary

Heroin scoring and injection in Melbourne AS (Dovey et al). The authors undertake a socio-

spatial analysis of injecting drug-use in public space. It focuses on one urban district in

Melbourne, Australia, which has become identified with heroin sale and use in public space.

Selling activities are camouflaged within a diverse street life while injecting sites are dispersed

through laneways, car parks and toilets. These injecting zones occupy liminal places that slide

between categories of private and public, and that mediate complex and paradoxical relations

between safety and danger. Those who inject in public space are caught in a dilemma of needing

both privacy and exposure in the event of an overdose; safety from police becomes danger from

an overdose. This empirical work, based on interview and spatial analysis, is presented as a basis

for theorizing the socio-spatial construction of heroin use and for assessing the prospects for safe

injecting

Notes

WEEK ONE: Dovey et al, 2001, melbourne smith street study

Case study is heroin scoring and shooting up on and around Smith street. Study area is a

down market strip of smith street. Has commercial and some industrial lots. Has back

alleys, parking lots etc .

Research Question: After purchase, where to heroin users go to shoot up? While some

users take the stuff home, there are street users; study focuses on street users and their

behaviour. Want to study drug use obtrusively, not allowing the users to know their, being

studied. To do this Dovey uses discarded heroin syringes as evidence.

Process Being Sampled: a process is a recurring behaviour that can be characterized by a

statistical population. A case is a person often being studied. Here the process is the

choice of location for heroin scorers to shoot up. Author simplifies reality, inner battle

between going far away to be out of sight, quick enough to use the drugs without them

being found on them, but public enough so people can find the subjects in case of

overdose.

Two questions arise, first is that are drug users rational\/ based on the addictive nature of

drug users, they do not seem to be displaying rational behaviour. Secondly, there isn't

much about street use shooting up that is specific to an individual.

Popular places for heroin shooting up included laneways, alcoves, public toilets and

parking lots.

Whose behaviour is being modelled? Here it is the heroin user' behaviour that is being

modelled, however other people (such as general public and police) affect the behaviour

of drug users.

Assumptions about behaviour: Author makes 4 assumptions in their study. Author

assumes heroin users are purposeful, assumes street life provides camouflage, assumes

drug user unsure about quality and overdose, assumes shooting where drug paraphernalia

is found.

Type of Study: Exploratory, there is no X variable being tested.

Nature of Data, Observation: Because the author wants to be obtrusive in their study, they

cannot view a person as a trial or an experiment. Instead, Dovey treats unofficial

personally outlined areas to find paraphernalia. In conventional application of scientific

method, 4 steps (model identification including theory, data included from lab

experiments including given and controlled situations, lab experiment and data, and

model validation including error analysis and hypothesis validation. All of this leads to

scientific method, something the authors did not do.

Often use survey or archival data, two steps to this; (1) exploration including literature

review, data assembly, model exploration and induction. (2) model validation including

statistical statistical methodology and hypothesis testing. Each observation then has a set

of conditions (X) that causes the outcome (Y).

Creation of the map untold, although assumes used through low altitude aerial photos and

existing governmental blueprints (archival). Used AM/FM or CAD techniques. Had to

use orthorectification in order to correct for distortion, altitude variation and curvature of

the earth.

Authors do not say what a zone is.

Statistical Population: Authors not sure what statistical population this study applies to.

Model: No model is used in this study.

Terms

Unobtrusively: they don’t want the behavior of users to be affected by their research project

Homogeneity: Homogeneity is a loose term to mean that the same model applies to every case in

the statistical population. However, this does not mean that every case has the same outcome. In

modeling, we imagine a deterministic part and a stochastic part. If two cases have the same level

for each independent variable, their deterministic parts will be the same. However, the notion of

a random variable, something that causes the level of Y for two observations to differ even if all

independent variable values are the same, explains why Y differs between them. Following the

treatment of a random variable, we typically imagine the stochastic component as having an

expected value of zero and a constant variance. The property that two observations can have

different realizations and yet have a stochastic component with a fixed variance is termed

homoscedasticity.

Risk: Probability, likelihood, or frequency with which an outcome occurs. Alternatively, the

proportion of cases that exhibit a common feature.

Exploratory date analysis - A field of study long associated with the work of John W Tukey, this

is the methodology of exploring data to discover relationships without imposing particular

mathematical structures and mindful of the presence of outliers. Topics range from visual

techniques (e.g., stem-and-leaf, box plots) to re-expression to plots of relationships and

straightening of plots to smoothing and polishing to robust estimation.

Feature: A generalization of the observed spatial pattern of entities. Typically, this includes a

cluster (hot spot), a hole, or a ridge on a map layer.

Y – layer entity - In GIS, the Y-layer maps the set of entities whose placement (SQ) or rate of

activity (RQ) we seek to explain.

Model identification: Procedures and methods by which the analyst selects the form of model

and the specifications of independent and dependent variables to be used in model estimation.

Model identification includes literature review, data assembly, model exploration, and induction.

Model validation: Procedures and methods by which the analyst assesses the appropriateness of

the selected model.

AM/FM - Automated mapping / facilities management. CAD applied at the level of buildings,

machinery, and fittings. A cousin to GIS software.

Null and alternative hypotheses: In science, falsification proceeds from the consideration of two

contradictory hypotheses about the population (process) under study: a null hypothesis, usually

exact (e.g., μ = 0), and an alternative hypothesis, usually inexact (e.g., μ&νβσπ; ≠ 0). The null

hypothesis can never be proven. A set of data can only reject a null hypothesis or fail to reject it.

Because of the nature of "empirical proof", we set up the null hypothesis as that we we aspire to

reject. The term "null hypothesis was first used in 1936 by the statistician Ronald Fisher. Jerzy

Neyman and Egon Pearson formalized the notion of the alternative hypothesis.

Zone (feature): A generalization of the observed spatial pattern of entities. Typically, this

includes a cluster (hot spot), a hole, or a ridge on a map layer.

Spatial context (geography of difference): GIS is useful only when there is a spatial context to

the analysis. In Economics and some other social sciences, the way that we think about problems

is often not spatial at all. However, for GIS to be useful, the phenomenon must be perceived in a

way that highlights the role of proximity and/or connectivity.

Spatial context takes two distinct forms. One form consists of variables that are independent of

the process under study. In studies of processes that are social, for example, we usually treat

physical variables as independent. Here, we would simply enter spatial context as one or more

independent variables into our model. The second form of spatial context is that variables may

not be independent of the process under study. Here, we have to be concerned about selection

bias.

Rate question: In a rate question (RQ), the analyst asks why the rate or level of some Y variable

varies from one place to the next. Why does a retail establishment at one site have higher sales

than an otherwise-similar store at another location? Why is the selling price of a house higher in

one neighbourhood compared to another. In a rate question, the dependent variable is typically,

but not always, ratio or interval-scaled. Another variant of a rate question is when we ask about

changes over time: e.g., survivorship. Of the retail establishments that were in the study area a

year ago, which ones are still here today? In this case, the dependent variable is categorical (e.g.,

1 if the store survives, 0 if not), not ratio or interval-scaled.

In the process under study, why does the likelihood, rate, intensity, or level of an outcome differ

from one observed location or area to the next? Rate question. emphasizes the importance of

spatial context. See also site question.