conclusions with this sample about online dating may not be prudent. This is even more relevant
when we realize that any finer dissection of the data may further reduce the observations.
dating$metonline <- factor(dating$metonline, levels = c("Met offline

With this knowledge in hand, I can now model median housing prices first as a function of
the lagged values of prices, followed by a model with two additional variables; that is, trend
and seasons (monthly dummies), and a third model with only the trend a

=
Prob Y j e
ei
z
j
Jz
1
ij
ij
Equation 9.25
For the conditional logit model, z i j = [ x i j ,w i ]. If x i j represents the attributes of the choices,
the subscript ij on x suggests that the x varies across the decision-makers ( i ) and choices ( j ),

I first determine the travel mode shares as they have been reported by the survey respondents:
27.6% respondents reported traveling by air, 14.3% reported traveling by bus, 28%
reported traveling by car, and 30% reported traveling by train (see Table 9.33

P walk
ln X X 2 t0 t1 1 t k k
Equation 9.19
g = 0 3 Equation 9.20
()()
()()()
()
()()=
+
=
+
P
g
ggg
g
gg
auto
exp
exp exp exp
exp
1 exp exp
1
123
1
12
Equation 9.21
()()
()()=
+
P transit
g
Community Type Rural 61.11 22.22 5.56 11.11
Suburban 56 24 10.67

Dependent Variable
Labor Force Participation
Kids under 6 1.463 *
(0.197)
Kids 6 to 18 0.065
(0.068)
Wifes age 0.063 *
(0.013)
Wife attended college 0.807 *
(0.230)
Husband attended college 0.112
(0.206)
Econometric Models of Binomial Data 361
Dependent V

family. An interesting question emerges about womens participation in the labor force when
they are not covered by their spouses health insurance. Would women lacking spousal coverage
be more likely to work full-time than do those with coverage?
I estimat

P auto
P transit
exp
exp
exp
exp
1
1
1
VV
VV
VV
VVta
ta
ta
a t Equation 9.34
Equation 9.35 gives the log of odds:
()()
()
= ( ) = P auto
P transit
ln ln exp V V V V
at
a t Equation 9.35
If I include a third choice, that is, walk, with the utility functio

Toronto Is a City Divided into the Haves, Will Haves, and Have Nots
Published on October 30, 2014, after the mayoral elections in Toronto.
Its a tale of two cities: Toronto the rich, and Toronto the poor. 14 The citys rich have
elected a mayor, John Tory,

with transit ridership irrespective of how far or near the community is from downtown Manhattan.
In addition, low-income communities nearest to the downtown have longer commute times
than mid- or high-income communities. This relationship reverses with di

The histogram in the top-left corner presents the distribution of yearsmarried . The most
frequently occurring cohort comprises those married for 12 years or more. The histogram for
religiousness indicates that anti-religion individuals constituted a smal

Similarly, divorced individuals are also less likely to own iPhones and are more likely to use
regular phones. The widowed individuals are also much more likely to use regular phones than
any type of smart phones. Parents are more likely to use iPhones an

Trend 30.007* 425.156* 30.007*
February 2858.187* 1784.459 2858.187*
March 38.305 628.534 38.305
April 2325.829* 1839.916 2325.829*
May 13.858 630.144 13.858
June 1174.273 45.858 1174.273
July 19.409 1.00E+03 19.409
August 863.909 1.20E+03 863.909
Septemb

p_singmom Percent of single-mother households
res_value Average self-reported house price
p_rental Percent of rental units
p_vacant Percent of vacant units
p_poverty Percent of households below the poverty line
p_black Percent of African American househol

at the location, and the mode of travel. The data collected in trip diaries was stored in a spatial
database using GIS software.
The analysis of the spatial database generated maps of tourists trip destinations in Hong
Kong, which helped reveal trends tha

ratio of odds (Prob:Work full-time/Prob:Unemployed) for working full-time against not working
when Spouses health insurance status equals 1.
The odds of wives covered by the spouses health insurance working full-time (against
not working) are e .701 = 0.4

takes less than a second to perform often takes significantly longer in QGIS. However, as a freeware
tool, it is stable and extremely useful.
Later in this chapter, I illustrate the use of Geoda, a freeware option for advanced spatial
econometrics.
GIS Da

484 Chapter 11 Doing Serious Time with Time Series
This becomes
Substituting = 1 in the preceding equation returns the following:
y y t t t 1 t
The modified version of the ADF test is shown in Equation 11.19 :
Equation 11.19
The null hypothesis states

+
=
+
=
Prob Y j e
e
ee
eei
xw
j
Jxw
xw
j
Jxw
11
ij j i
ij j i
ij j i
ij j i
'
'
Equation 9.42
where x ij are the alternative-specific covariates, whereas w i are individual-specific attributes.
One can include as a regressor an individual-specific varia

predict resar101, res
wntestb resar101, table / p>.05, residuals not different from white
noise
arima to_st_to, arima(2,0,0)
est store ar200
predict ar200, y
label variable ar200 "arima(2,0,0)"
arima to_st_to, arima(2,0,1)
est store ar201
predict ar201, y

Equation 11.29
Note that y t is a function of distributed lags of past values of x . The coefficients on lagged
x s are called Lagged Weights whereas their pattern is called the Lag Distribution. The preceding
equation is estimated as an OLS model where l

Assume that we perform PCA on 10 variables, which were normalized to mean = 0, and
variance =1. Lets assume that the total variance explained by the first eigenvector is 4.0. Because
the total number of variables is 10, we can calculate 410 * 100 to concl

starts are one of the most watched economic indicators because they are considered a leading
indicator of changes in economic production. Housing starts decline in advance of a recession,
and rise before post-recession economic recovery ensues.
Figure 11.

The incidence of lower household income bears a positive correlation with higher incidence
of rental housing units. This is depicted in Figure 10.33 , which shows that the neighborhoods
with high concentration of African-American households report a highe

Given the ubiquitous use of smart phones by the youth in cities, one may be forgiven to assume
that such trends are common elsewhere and among all age cohorts.
The battle for smart phone consumers has resulted in tremendous innovation in product and
servi

Second is the implied assumption that growing income inequality by default is at odds with
social cohesion and long-term economic and social viability of the region. This assumes that the
social safety nets, which are predominantly supported by the taxes