In this paper, we characterize the structure and topological indices of a class of random spider trees (RSTs) such as degree-based Gini index, degree-based Hoover index, generalized Zagreb index and other indices associated with these. We obtain the exact and asymptotic distributions of the number of leaves via probabilistic methods. Moreover, we relate this model to the class of RSTs that evolves in a preferential attachment manner.
For n \geq 1 and 0 < p < 1, the following statements hold:
\mathbb{E}(L_{n})=3+(n-1)p and \mathbb{V}(L_{n})=(n-1)(1-p)p.
M_{L_{n}}(t)=(1-p+pe^{t})^{n-1}e^{3t}, t \in \mathbb{R}.
For each k \in \mathbb{R}, \displaystyle\frac{L_{n}-3-(n-1)p}{\sqrt{p(1-p)(n+k)}} \xrightarrow{D} N(0,1) as n \rightarrow \infty.
Let TI_{n} the value of the topological index at stage n. For each n \geq 1, we have
\mathbb{E}(TI_{n})=\mathbb{E}(h(L_{n})^{\alpha})+(h(1)^{\alpha}-h(2)^{\alpha})\mathbb{E}(L_{n})+h(2)^{\alpha}(n+2) and
\mathbb{V}(TI_{n})=\mathbb{V}(h(L_{n})^{\alpha}+(h(1)^{\alpha}-h(2)^{\alpha})L_{n}).
Note that I_{n}+L_{n}+1=n+3, where I_{n} is the number of internal nodes in the tree at stage n, it follows that:
TI_{n}= h(L_{n})^{\alpha}+h(1)^{\alpha}L_{n}+h(2)^{\alpha}(n+2-L_{n})
By (Proof), we immediately get the mean and the variance of TI_{n}.
If h(L_{n})=aL_{n}+b with a,b \in \mathbb{R} then M_{h(L_{n})}(t)= (1-p+pe^{at})^{n-1}e^{(3a+b)t}, t \in \mathbb{R} and n \geq 1.
Let \alpha \in \{1,2,\dots\} and t \in \mathbb{R}, we have
\displaystyle\frac{d^{\alpha} M_{L_{n}}}{dt}(t)=\displaystyle\sum_{i=1}^{\alpha}C_{\alpha,i}\hspace{0.05cm}p^{i}e^{it}\displaystyle\frac{d^{i} M_{L_{n}}}{du}(u(t)),
with C_{\alpha,\alpha}=C_{\alpha,1}=1 and C_{\alpha,i}=C_{\alpha-1,i-1}+iC_{\alpha-1,i} for i\in \{2,3, \dots ,\alpha -1\}.
We will get the proof via mathematical induction on \alpha. Firstly observe that, Proposition 1 b may be simplified by defining a new variable u(t)=1-p+pe^{t}. Substituting 1-p+pe^{t} by u(t) we get M_{L_{n}}(t)=M_{L_{n}}(u(t))=u(t)^{n-1}(\frac{u(t)+p-1}{p})^{3}. Thus, for the base, that is \alpha =3, we have
\displaystyle\frac{d^{3} M_{L_{n}}}{dt}(t)=p^{3}e^{3t}\displaystyle\frac{d^{3} M_{L_{n}}}{du}(u(t))+3p^{2}e^{2t}\displaystyle\frac{d^{2} M_{L_{n}}}{du}(u(t))+pe^{t}\displaystyle\frac{d M_{L_{n}}}{du}(u(t)),
with C_{3,1}=C_{3,3}=1 and C_{3,2}=C_{2,1}+2C_{2,2}=3. We assume that the statement holds for all \alpha, i.e.
with C_{\alpha,\alpha}=C_{\alpha,1}=1 and C_{\alpha,i}=C_{\alpha-1,i-1}+iC_{\alpha-1,i} for i\in \{2,3, \dots ,\alpha -1\}. Note that for each i\in \{1,2, \dots ,\alpha\} we arrive at
\displaystyle\frac{d }{dt}\left( C_{\alpha,i}p^{i}e^{it}\displaystyle\frac{d^{i} M_{L_{n}}}{du}(u(t))\right)
By (?) and (?), we have proved the following result
with C_{\alpha+1,\alpha+1}=C_{\alpha+1,1}=1 and C_{\alpha+1,i}=C_{\alpha,i-1}+iC_{\alpha,i} for i\in \{2,3, \dots ,\alpha\}, which completes the proof.
For \alpha \in \{1,2,\dots\}, it is verified that
\displaystyle\frac{d^{\alpha} M_{L_{n}}}{dt}(0)=\displaystyle\sum_{i=1}^{\alpha}C_{\alpha,i}\hspace{0.05cm}p^{i}\displaystyle\frac{d^{i} M_{L_{n}}}{du}(1),
with C_{\alpha,\alpha}=C_{\alpha,1}=1 and C_{\alpha,i}=C_{\alpha-1,i-1}+iC_{\alpha-1,i} for i\in \{2,3, \dots ,\alpha -1\}.
For n\geq 1, p \in (0,1) and \alpha \in \{1,2,\dots\} the following identity holds
\displaystyle\frac{d^{\alpha} M_{L_{n}}}{dt}(0)= p^{\alpha}n^{\alpha}+\frac{\alpha}{2}(\alpha(1-p) -p+5)p^{\alpha-1}n^{\alpha-1}+O(n^{\alpha-2}).
First, observe that
M_{L_{n}}(u(t))=\frac{1}{p^{3}}(u(t)^{n+2}+3(p-1)u(t)^{n+1}+3(p-1)^{2}u(t)^{n}+(p-1)^{3}u(t)^{n-1}),
t \in \mathbb{R}. Then for each i \in \{1,2,\dots\}, we obtain
\displaystyle\frac{d^{i} M_{L_{n}}}{du}(u(0))=\displaystyle\sum _ {k=-1}^{2}\frac{b_{k}}{p^{3}}(n+k)(n+k-1) \ldots (n+k - (i-1))u(0)^{n+k-i},
with b_{-1}=(p-1)^{3}, b_{0}=3(p-1)^{2}, b_{1}=3(p-1) and b_{2}=1.
Accordingly, it follows that
\displaystyle\frac{d^{i} M_{L_{n}}}{du}(1)=\frac{1}{p^{3}}\displaystyle\sum _ {k=-1}^{2}b_{k}n^{i}+\frac{1}{2p^{3}}\displaystyle\sum _ {k=-1}^{2}b_{k}i(2k+1-i)n^{i-1}+ O(n^{i-2})
~~~~~~~~~~~~~~~~~~=n^{i}-\frac{i(ip+p-6)}{2p}n^{i-1}+ O(n^{i-2}).
By Corollary 1, it is concluded that for each \alpha \in \{1,2,\dots\}
\displaystyle\frac{d^{\alpha} M_{L_{n}}}{dt}(0)= p^{\alpha}n^{\alpha}-\frac{\alpha}{2p}(\alpha p+p-6)p^{\alpha}n^{\alpha-1}+C_{\alpha,\alpha -1}\hspace{0.05cm}p^{\alpha-1}n^{\alpha-1}+O(n^{\alpha-2}),
as C_{\alpha,\alpha-1}=C_{\alpha-1,\alpha-2}+(\alpha-1)C_{\alpha-1,\alpha-1}=C_{\alpha-1,\alpha-2}+\alpha-1=\frac{\alpha(\alpha-1)}{2}. Then
\displaystyle\frac{d^{\alpha} M_{L_{n}}}{dt}(0)= p^{\alpha}n^{\alpha}+\frac{\alpha}{2}(\alpha(1-p) -p+5)p^{\alpha-1}n^{\alpha-1}+O(n^{\alpha-2}).
For any \alpha \in \{3,4, \dots\}, it is verified that \frac{Z^{g}_{n}}{n^\alpha}\xrightarrow{P}p^{\alpha}, when n goes to infinity.
Let X_{n}=\frac{1}{n^\alpha}(Z^{g}_{n}-\frac{\alpha}{2}(\alpha(1-p) -p+5)p^{\alpha-1}n^{\alpha-1}-O(n^{\alpha-2})) by Chebyshev’s inequality [17] we get
\mathbb{P}(|X_{n}-p^{\alpha}|\geq \epsilon) \leq \frac{1}{n^{2\alpha}\epsilon^{2}}(\alpha^{2}(1-p) p^{2\alpha-1}n^{2\alpha-1}+O(n^{2\alpha-2})),
for any \epsilon >0. If n \rightarrow \infty then \frac{1}{n^{2\alpha}\epsilon^{2}}(\alpha^{2}(1-p) p^{2\alpha-1}n^{2\alpha-1}+O(n^{2\alpha-2})) \rightarrow 0, so X_{n} \xrightarrow{P} p^{\alpha}. On the other hand, X_{n}=\frac{Z^{g}_{n}}{n^\alpha}-x_{n} with x_{n}= \frac{1}{n^\alpha}(\frac{\alpha}{2}(\alpha(1-p) -p+5)p^{\alpha-1}n^{\alpha-1}+O(n^{\alpha-2})) then x_{n} \rightarrow 0 as n\rightarrow \infty. Therefore, the result follows.
For any \alpha \in \{3,4, \dots\} and r > 0, it is verified that \frac{Z^{g}_{n}}{n^\alpha}\xrightarrow{L_{r}}p^{\alpha}, when n goes to infinity.
Using Theorem 2 we have \frac{Z^{g}_{n}}{n^\alpha}\xrightarrow{P}p^{\alpha} when n goes to infinity. On the other hand, \frac{Z^{g}_{n}}{n^\alpha} \geq 0 for all n\geq 1, then |\frac{Z^{g}_{n}}{n^\alpha}|=\frac{Z^{g}_{n}}{n^\alpha}. For each r > 0 there exists N \in \{1,2, \dots \} such that N > r. By Theorem 4.2 in Chapter 5 of [17] we obtain \{\left(\frac{Z^{g}_{n}}{n^\alpha}\right)^{r},~ n \geq 1\} is uniformly integrable. Then, applying Theorem 5.4 in Chapter 5 of [17] with X_{n}=\frac{Z^{g}_{n}}{n^\alpha}, we obtain the convergence in r-mean and the proof is completed.
For p \in (0,1), when n goes to infinity it is verified that
For all k \in \mathbb{R}, \displaystyle\frac{Z_{n}-n^{2}p^{2}}{2\sqrt{p^{3}(1-p)(n+k)^{3}}} \xrightarrow{D} N(0,1).
For all r >0, \frac{Z_{n}}{n^2}\xrightarrow{L_{r}}p^{2}.
a) By (?) Z_{n}=\left(L_{n}-\frac{3}{2}\right)^{2}+\frac{16n+23}{4} for each n \geq 1. Note that Proposition 1 c implies that \displaystyle\frac{L_{n}-\frac{3}{2}}{\sqrt{p(1-p)(n+k)}} is equivalent to a normal random variable with mean \displaystyle\frac{\frac{3}{2}+(n-1)p}{\sqrt{p(1-p)(n+k)}} and variance 1 in distribution. Therefore, \left( \displaystyle\frac{L_{n}-\frac{3}{2}}{\sqrt{p(1-p)(n+k)}}\right)^{2} \sim \chi^{2} \left(\displaystyle\frac{\left(\frac{3}{2}+(n-1)p\right)^{2}}{p(1-p)(n+k)},1\right). Indicating us that:
\displaystyle\frac{Z_{n}}{p(1-p)(n+k)}-\displaystyle\frac{16n+23}{4p(1-p)(n+k)} \sim \chi^{2} \left(\displaystyle\frac{\left(\frac{3}{2}+(n-1)p\right)^{2}}{p(1-p)(n+k)},1\right).
By the well-known normal approximation of non-central chi-squared distribution (see [11]), it is obtained that
\displaystyle\frac{\displaystyle\frac{Z_{n}}{p(1-p)(n+k)}-\displaystyle\frac{16n+23}{4p(1-p)(n+k)}-(1+\displaystyle\frac{\left(\frac{3}{2}+(n-1)p\right)^{2}}{p(1-p)(n+k)})}{\sqrt{2(1+\displaystyle\frac{2\left(\frac{3}{2}+(n-1)p\right)^{2}}{p(1-p)(n+k)})}} \xrightarrow{D} N(0,1),
as n \rightarrow \infty. In particular,
\displaystyle\frac{\displaystyle\frac{Z_{n}}{p(1-p)(n+k)}-\displaystyle\frac{16n+23}{4p(1-p)(n+k)}-1-\displaystyle\frac{\left(\frac{3}{2}+(n-1)p\right)^{2}}{p(1-p)(n+k)}}{\sqrt{2(1+\displaystyle\frac{2\left(\frac{3}{2}+(n-1)p\right)^{2}}{p(1-p)(n+k)})}}
~~~~~~~~~~~=\displaystyle\frac{Z_{n}-\frac{16n+23}{4}-p(1-p)(n+k)-\left(\frac{3}{2}+(n-1)p\right)^{2}}{\sqrt{p(1-p)(n+k)}\sqrt{2(p(1-p)(n+k)+2\left(\frac{3}{2}+(n-1)p\right)^{2})}}
=\displaystyle\frac{Z_{n}-\frac{16n+23}{4}-p(1-p)(n+k)-\left(\frac{3}{2}+(n-1)p\right)^{2}}{\sqrt{4p^{3}(1-p)(n+k)^{3}+o(n^{3})}}
=a_{n}\displaystyle\frac{Z_{n}-n^{2}p^{2}}{\sqrt{4p^{3}(1-p)(n+k)^{3}}}+b_{n},
where
a_{n}=\displaystyle\frac{\sqrt{4p^{3}(1-p)(n+k)^{3}}}{\sqrt{4p^{3}(1-p)(n+k)^{3}+o(n^{3})}},
b_{n}=\displaystyle\frac{o(n^{\frac{3}{2}})}{\sqrt{4p^{3}(1-p)(n+k)^{3}+o(n^{3})}}.
It is verified that a_{n} \rightarrow 1 and b_{n} \rightarrow 0 as n \rightarrow \infty, then
\displaystyle\frac{Z_{n}-n^{2}p^{2}}{2\sqrt{p^{3}(1-p)(n+k)^{3}}} \xrightarrow{D} N(0,1).
b) The proof can be done similarly to that of Corollary 2.
For p \in (0,1), when n goes to infinity it is verified that
For all k \in \mathbb{R}, \displaystyle\frac{S_{n}-\frac{n^{2}p^{2}}{2}}{\sqrt{p^{3}(1-p)(n+k)^{3}}} \xrightarrow{D} N(0,1).
For all r>0, \frac{S_{n}}{n^{2}}\xrightarrow{L_{r}}\frac{p^{2}}{2}.
For p \in (0,1), when n goes to infinity it is verified that
For all k \in \mathbb{R}, \displaystyle\frac{P_{n}-n^{2}p^{2}}{2\sqrt{p^{3}(1-p)(n+k)^{3}}} \xrightarrow{D} N(0,1).
For all r >0, \frac{P_{n}}{n^{2}} \xrightarrow{L_{r}} p^{2}.
For all r >0 it is verified that, \frac{F_{n}}{n^{3}}\xrightarrow{L_{r}}p^{3} as n \rightarrow \infty.
As n \rightarrow \infty, we have \mathbb{E}(G_{n}) \rightarrow \frac{p(2-p)}{2}.
It is verified that G_{n}\xrightarrow{P} \frac{p(2-p)}{2}, as n \rightarrow \infty.
By Chebyshev’s inequality [17] we have
\mathbb{P}(|G_{n}-\mathbb{E}(G_{n})|\geq \epsilon) \leq \frac{\mathbb{V}(G_{n})}{\epsilon^{2}},
for any \epsilon > 0. If n \rightarrow \infty then \frac{\mathbb{V}(G_{n})}{\epsilon^{2}} \rightarrow 0, so G_{n}-\mathbb{E}(G_{n}) \xrightarrow{P} 0. Therefore, Proposition 8 completes the proof.
For all r >0, we have G_{n}\xrightarrow{L_{r}} \frac{p(2-p)}{2}, when n goes to infinity.
For n \geq 1, we have
H_{n}=\frac{(n+1)L_{n}}{2(n+3)(n+2)}.
\mathbb{E}(H_{n})=\frac{ pn^{2}+3n+3-p}{2(n+3)(n+2)}.
\mathbb{V}(H_{n})=\frac{p(1-p)(n^{2}-1)(n+1)}{4(n+3)^{2}(n+2)^{2}}.
For all r >0, H_{n}\xrightarrow{L_{r}} \frac{p}{2}, as n \rightarrow \infty.
In the preferential model for all r >0, it is verified that H_{n} \xrightarrow{L_{r}} \frac{1}{4} and \mathbb{E}(H_{n}) \rightarrow \frac{1}{4} when n goes to infinity.
The text on this page is from arXiv:2204.10840v1 by Saylé C. Sigarreta, Saylí M. Sigarreta and Hugo Cruz-Suárez, under the licence CC BY 4.0. The corpus made these changes: it read the items out of the LaTeX source, it expanded the macros of the paper, it put each proof after its statement, and it converted the text to HTML. The formulas are the TeX of the paper, and KaTeX renders them. A formula or a reference can look different from the PDF. The page is published under CC BY 4.0.
The record of each item is in analysis/objects/2204/2204.10840.jsonl.