I’m giving a talk to the Statistics students at the University of Central Missouri. Here’s the title and abstract of my talk.

Recent (and not so recent) advances in statistical computing.

Like many other disciplines, Statistics has benefited from Moore’s Law. As computers have become smaller, faster, and cheaper, statisticians have increased the capability of their data analysis tools. This talk reviews some of these computer dependent advances over the last 40 years.

I’m going to place some preliminary notes here to help me develop my Powerpoint slides.

Moore’s Law is invoked in many different forms, but it was originally derived in 1965 by Gordon Moore, co-founder of Intel Corporation. It states that the numer of components of an integrated chip doubles every year, every 1.5 years, or every two years. The exact form is less critical than the fact that this represents an exponential growth. As the number of circuits in an integrated circuit increases, the speed increases more or less proportionately.

So, for example, the 8088 processor, the heart of the original IBM PC, was introduced in 1979, had 29,000 transistors and could calculate 0.3 million instructions per second (MIPS). The 80286 chip, introduced in 1982, had 139,000 transistors and could calculate 1.2 MIPS. The 80386 chip, introduced in 1985, had 275,000 transistors and could calculate 5.5 MIPS. The 80486 chip, introduced in 1989, had 1.2 million transistors and could calculate 20 MIPS. The first Pentium chips were introduced in 1993, had 3.1 million transistors and could calculate 100 MIPS.

Why is this important? Consider the estimation of regression coefficients in a linear regression model.