The explanation obtained, by the statistical analysis, 
shows some significant and insignificant behaviors. 
• Indeed, taking into account a certain type of 
activities (clothes, shoes, etc), the flow of clients 
depends strongly on some other variable. There is a 
linear dependence on the surface of the business 
activity. Moreover there is a linear dependence on 
the inverse of the square of the distance from the 
main street. 
• On the contrary, taking into account a completely 
different type of activities (food), the flow of clients 
is quite independent on some other variable. There is 
a linear dependence on the sinus of both the surface 
of the business activity and its distance from the 
main street. 
This kind of explanation has been considered reasonable 
and satisfactory, because the first class of business 
activities are expected to be sensitive to the other 
variables, whilst the second class seems to be quite 
insensitive. 
4. A TWO WAY VARIANCE ANALYSIS 
The last part of the spatial analysis has been only a 
simulation. Indeed although space - time depending data 
analysis is very important, to understand the behavior of 
many phenomena and processes, referred to the space 
and rapidly changing in the time, data acquisition is often 
really very, very expensive. 
Therefore an old database, subdivided into regions and 
epochs, has been acquired and it has formed the set of 
data, collected together into two - dimensional cells. The 
information into the cells is represented by a statistical 
mono - dimensional variable whose elements are the 
values of the attributes of the information itself. 
A more sophisticated investigation could be done, by 
using a three - way variance analysis, to better exploit the 
spatial variability. Furthermore a statistical multi - 
dimensional variable increases by the repetitions, as 
much as possible, the degrees of freedom. Finally the 
interactions among the cells themselves can be also taken 
into account. 
In the present work, a two - way variance analysis 
without repetitions has been done. The different areas of 
the city represent the blocks and the different epochs of 
collected data are the treatments. 
*** Two-way Variance of Analysis *** 
Legend: VI = Blocks 
V2 = Treatments 
VI 
V2 
Residuals 
A 
23768.17 
2365.42 
20038.42 
B 
107 
5 
535 
C 
222.13 
473.08 
37.45 
D 
5.93 
12.63 
F 
3.06 
1.42 
Note: The estimated effects are balanced. 
The principal aim of the variance analysis is to maximize 
the layer - variance, describing the phenomena according 
to some selected the layer: rows, columns (piles) models, 
respect to the residual variance which, on the contrary, 
represents the random variability of the data. 
The two - way variance analysis shows that the estimated 
effects are balanced. This means that both the spatial 
variability and the time variability are significantly 
greater than the random errors. The simulation appears 
reasonable and satisfactory, fully confirming the studies 
performed analyzing real data. 
REFERENCES 
Barteime N.: Geoinformatik - Modelle, Strukturen, 
Funktionen. Springer Verlag, 1995. 
Carosio A.: Geoinformationssysteme'. ETH Zürich 
Department Geodätische Wissenschaften Zürich 1997. 
Carosio A.: Fehlertheorie und Ausgleichung rechnung. 
ETH Zürich Department Geodätische Wissenschaften 
Zürich 1998. 
Everitt BrianS.: Statistical Analysis using S-Plus. 
Chapman&Hall 1996. 
Flury B., Riedwyl H.: Angewandte multivariate Statistik. 
Gustav Fischer Verlag 1983. 
Morrison D.F.: Multivariate Statistical Methods. Mc 
Graw Hill 1989. 
A = Sums of the Squares 
B = Degrees of Freedom 
C = Variances 
D = F of Fisher (expected values) 
E = F (alpha = 1%, on two sides)