### Table 4. Comparison of shape distributions versus moments.

"... In PAGE 11: ... The shape signatures are compared using a component-by-component L2 norm. Table4 compares the results achieved with this moment- based classifier versus the method proposed in this paper. We find that D2 shape distributions outperform moments for classification of models in our tests.... ..."

### Table 10: Conjugated gradients on a nite di erence discretization, test1.i We test the multigrid method with the given restriction and prolongation stencils, see 8 les are in MGfdm7/Verify/

### Table 1 Comparison of descriptive statistics for continuous and corresponding discrete Rayleigh distributions with parameters chosen to match their means

1998

"... In PAGE 5: ... 2 compares the shapes of the discrete and continuous Rayleigh distributions. Table1 com- pares the descriptive statistics for the continuous and corresponding discrete Rayleigh distribu- tions with parameters chosen to match their means. It is evident that the discrete Rayleigh distribution approximates the Rayleigh distribu- tion.... ..."

### Table 1 Characteristics of Discrete Choice Models

1998

"... In PAGE 4: ...3 We also discuss the common assumptions these two logit models make about the distribution of the disturbances. We characterize the basic properties of several discrete choice models in Table1 .[Table 1 Here] Table 1 shows that multinomial logit is the most restrictive discrete choice model we discuss in this paper; it models the choice probabilities as functions only of characteristics of the individual voter, it does not allow the error terms to be correlated across choices, and it provides few answers to important political questions.... In PAGE 4: ... We characterize the basic properties of several discrete choice models in Table 1.[Table 1 Here] Table1 shows that multinomial logit is the most restrictive discrete choice model we discuss in this paper; it models the choice probabilities as functions only of characteristics of the individual voter, it does not allow the error terms to be correlated across choices, and it provides few answers to important political questions. Conditional logit, however, is less restrictive since it allows for... ..."

Cited by 17

### Table 2. Generators for discrete distributions

1992

"... In PAGE 5: ...his yields convenient algorithms with bounded computation times, i.e. the times do not grow with the parameters of the distributions. Again unstored or simple inversion is substituted in the utility routines pruec, bruec and hruec when it is faster than ratio of uniforms (see Table2 below). The geometric distribution with probability of success p and probability function pk = p (1 ? p)k; k = 0; 1; : : : permit inversion by K = bln U=(1?p)c .... ..."

Cited by 5

### Table 5: Discretization errors on meshes with irregular-shaped polyhedra.

in The mimetic finite difference discretization of diffusion problem on unstructured polyhedral meshes

2005

"... In PAGE 17: ... But now we investigate the dependence of convergence rates on shape regularity of mesh elements. The computational results are presented in Table5 where is the ratio of the maximal face area to the minimal one. Thus, = 1 means that a hexahedron is transformed into a pentahedron.... In PAGE 17: ...Table 5: Discretization errors on meshes with irregular-shaped polyhedra. The numerical results presented in Table5 verify that decrease of shape regularity of mesh elements does not a ect convergence of the mimetic discretization. 6.... ..."

Cited by 2

### Table 11b: Model Restriction Test Model Restriction Test

2002

Cited by 10

### Table 5: Testing Demand Restrictions

"... In PAGE 17: ... An important issue is whether the restrictions implied by demand theory hold. Likelihood ratio tests shown in Table5 indicate that for the LA/AIDS demand system, symmetry and homogeneity are individually rejected, and, as a result, both symmetry and homogeneity to- gether do not hold. On the other hand, there is strong support for accepting symmetry for the Rotterdam specification and limited support for accepting the null that homogeneity and homogeneity and symmetry together hold.... ..."

### Table 3: Test of cointegrating restrictions.

### Table 2. Statistics for separate color and shape

"... In PAGE 5: ...ompleting a query. The tests shown are referred to combined color and shape retrieval. Table 1 shows that building a single tree for multiple features requires less time than building a tree for every feature. Considering the queries, combined features #28Table 3#29 are faster than separate features #28 Table2 #29 because in the latter case it is necessary to search each index and then merge the results. We notice that for larger values of k, the performance of separate features vs.... In PAGE 7: ... 3. Complexity of merge operation Table2 and Figure 2 show that both in the real dataset of 400 images and in randomly generated distributions the metric which provides better performance is L 1 , then L 2 and #0Cnally L 1 ; however, performance is inversely proportional to the retrieval accuracy, which is better with L 1 , then with L 2 and #0Cnally L 1 metric. Figure 3 veri#0Ces, for the merge operation, the computational complexity estimated in Section 4, which is polynomial on the number n of features requested by the query and the number k of images to be retrieved.... ..."