The Dictionary of Human Geography

illness or illness causing unemployment.
Measurements over time usually show strong
state dependence in that individuals do not
move rapidly and continually between differ-
ent states, so that current behaviour is influ-
enced by past or previous behaviour. Only
LDA is capable of taking account of prior
information when examining current situ-
ations. Moreover, only longitudinal data can
separate age andcohort effects – the life
experiences of those aged over sixty years
may be quite different between cohorts born
before and after 1945. Unlike cross-sectional
analysis, which analyses variation between
cases, LDA also works within cases between
occasions. As such, it is much better able to
take account of ‘unobserved heterogeneity’,
unexplained variation due to the omission
of explanatory variables that are either un-
measured or even un-measurable. LDA will
improve control for such heterogeneity and
help to provide a measure of the extent of its
presence (Davies and Pickles, 1985).
Longitudinal data sets have a number of
features that provide a challenge for analysis.
Thus if the outcome has not yet occurred the
sequence is censored, so that in an analysis of
longevity there will be people who are still alive
at the end of observation period. Standard stat-
isticalmodelsare based on the assumption of
independence, but repeated measures over
time are likely to be strongly autocorrelated.
Missing data are also a particular problem, as
the requirement of multiple follow-up often
leads to attrition: this is a particularly challen-
ging problem when the drop-out is informative,
because it depends on what would have been
observed if the person had not dropped out.
Technically, there are three broad sets of
approaches to LDA:

 Repeated measures analysis, in which data
on repeated occasions are seen as being
nested within individuals, so that we can
examine, for example, the growth of in-
come over time and evaluate the changing
gender gap;multi-level modelling is
increasingly being used for such data, as
it does not require the measurement of
every individual on all occasions to allow
between-individual, within-individual and
between-occasion variation to be explicitly
modelled.
 Event history analysis(duration analysis or
hazard modelling) is concerned with the
timing of transitions from one state to
another, so that it is possible to test how
the transition rate of moving from one state

(e.g. unemployed) to another (employed)

is affected by other variables such as edu-

cation level: these explanatory variables

may be either time invariant (such as

gender) and/or time varying. In such

analyses, the dependent variable is the

duration until event occurrence. Single

non-repeatable event analysis (e.g. the

transition to death) was developed first

and is often called ‘survival analysis’, but

it is now possible to analyse repeated

events (e.g. multiple spells of unemploy-

ment and employment), competing risks

(e.g. different reasons for leaving a job

such as voluntary choosing another job,

redundancy, retirement), multiple states

(e.g. transitions between single, marriage

and cohabitation states) and multiple

processes (e.g. joint modelling of partner-

ship and employment histories).

 sequence analysis, which works holistic-

ally to identify characteristic time-based

trajectories.

These methods have been extended to model

spatial choice (Wrigley, 1990), the duration

of point patterns (Pellegrini and Reader,

1996) and spatial processes more generally

(Waldorf, 2003). kj

Suggested reading

Allison (1984); Blossfeld and Rohwer (2002);

Dale and Davies (1994); Fitzmaurice, Laird

and Ware (2004); Ruspini (2002); Singer and

Willett (2003); Taris (2000); Wrigley (1986).

Los Angeles School A loose affiliation of Los

Angeles-based geographers and urbanists, con-

nected by their theorizations of the LA region

and the implications of its dynamics for all cit-

ies. The School is associated with a notion of

‘postmodern urbanism’ and an argument that

LA’s complex urbanform, new economicclus-

tersandinequalitysignalthefutureofurban-

ism(seepostmodernism; postmodernity). Its

members argue that contemporary urbanism

can be understood less through the concepts

of thechicago schooland more through the

fragmentary geography of Southern California

(Dear, 2001). The School’s work has been

the subject of much criticism and debate con-

tinues on the efficacy of framing LA, or any

single city, as an exemplar of urbanism (Dear

and Dahmann, 2008; Mollenkopf, 2008;

Simpson and Kelly, 2008). em

Suggested reading

Dear (2001).

Gregory / The Dictionary of Human Geography 9781405132879_4_L Final Proof page 432 31.3.2009 2:44pm

LOS ANGELES SCHOOL