Why Robert Bartle excluded the Riemann-Stieltjes integral in the last edition?

221 Views Asked by At

The question is about "The elements of real analysis" by Robert Bartle. In 1st and 2nd editions there are chapters on the Riemann-Stieltjes integral. But in 3rd and 4th editions he replaced it by the Riemann integral.

The reason of the question is the following one : At first I started to study the 2nd edition because I did not know that there are more recent editions. Now I know it. The R.-S. integral is more general than the Riemann integral. So what I must to study?

New editions must be "better". (I think)

So what is the reason for this replacing and is there any need to read old editions?

EDIT: In the introductory chapters in 3rd and 4th editions I did not find the explanation for this.

EDIT: I see that there is an attempt to close this question because someone thinks it is opinion-based. I dont know. I think this question must have resonanle answer. Something like: "There is gauge integral in the reference in the 4th edition. And this integral is more general than R-S integal. R-S is old-fashion now." It is just my fantasy. I want to say that this question must have some meaningfull answer.

1

There are 1 best solutions below

4
On BEST ANSWER

I agree with @Masacroso ... Now that measure theory is commonly studied, the Riemann-Stieltjes integral is not needed.

Example: In former times the R-S integral was seen in probability theory. $$\mathbb{E}[\phi(X)] = \int_{\mathbb{R}}\phi(t)dF(t)$$ where function $F$ is the CDF of the random variable $X$. Nowadays we write $$\mathbb{E}[\phi(X)] = \int_{\mathbb{R}}\phi(t)d\mu(t)$$ where measure $\mu$ is the distribution of the random variable $X$ in the sense of $\mu(E) := \mathbb{P}(X\in E)$.


There may still be some uses for Stieltjes-type integrals. But not enough to justify including them in a course that every mathematician takes.

$\bullet\quad$ Stochastic integrals,
$$ \int_0^t X_s\;dY_s $$ where $(X_s)$ and $(Y_s)$ are semimartingales with respect to a stochastic basis $(\mathcal{F}_s)$. Defined as a limit (in mean) of Riemann sums.

$\bullet\quad$ The spectral theorem: if $T$ is a self-adjoint operator on Hilbert space $H$, $$ T = \int_{\sigma(T)} \lambda\;dE_\lambda $$ where $\lambda \mapsto E_\lambda$ is a certain nondecreasing family of projections on $H$. The (measurable) functional calculus is then formulated $$ \phi(T) = \int_{\sigma(T)} \phi(\lambda)\;dE_\lambda $$ where $\phi$ is a measurable function defined on $\sigma(T)$.