Appendix AB. Who uses this book and its associated web site, documents and software?

In the last few years, this site (http://terpconnect.umd.edu/~toh/spectrum/) has been accessed from Internet Service Providers in over 162 countries and 6 non-region-specific categories (e.g. satellite providers), including many countries in the developing world, some very small countries (e.g. Liechtenstein, the Faroe Islands), relatively isolated countries (Cuba, North Korea, Myanmar/Burma), and even some war-torn regions (Afghanistan, Syria, Iraq). Breath of Internet access is often an issue. For example, I've got fewer views from Cuba that from other Spanish-speaking countries with smaller populations, such as Bolivia, Dominican Republic, Costa Rica, Puerto Rico, Panama, and Uruguay, even though Cuba has many active scientists, especially in the medical and pharmaceutical fields.

The first Web version went up in 1996, but I didn't start keeping track of views until 2008; since then there have been over 2 million page views. The distribution of page view counts among countries is very long-tailed, with one-third of the views coming from the USA (except during major US holidays), half of the views coming from only 5 countries (USA, India, Germany, United Kingdom, and China) and 99% of the views coming from only 39 countries. Among the countries that have a relatively large number of page views relative to their populations are the USA, Germany, UK, Canada, Australia, Netherlands, Switzerland, Singapore, Israel, Belgium, Taiwan, South Korea, and Scandinavia. (Another web site of mine on a related subject, Interactive Computer Models for Analytical Chemistry Instruction, had got an additional 825,000 views).

The Internet Service Providers with the largest number of views are Comcast, Verizon FIOS, Time Warner, Cloudflare, At&t U-verse, Deutsche Telekom (Germany), BSNL (India), and Cox Communication. Most views worldwide come from Windows machines, about 20% from Linux and Macintosh, and 10% from mobile devices. I've made efforts to make my pages more usable from mobile devices like smartphones.

About one quarter of the views come directly from educational institution ISPs that have "School", "Ecole", "College", "Hochschule", "Univ...", "Academic", or "Institute of Technology" in their names. (The number of educational users is certainly larger than that because some users are no doubt accessing from other ISPs in homes or businesses). An analysis of 200,000 views in 2015 showed that the biggest educational users have been the University of California System (UCLA, Berkley, etc.), Indian Institute Of Technology system, the University of Texas system, Massachusetts Institute Of Technology, the University of Michigan, the University of Maryland (my home institution), Delft University of Technology (Netherlands), Stanford University, China Education And Research Network Center, the University Of Wisconsin System, and the University of Illinois.

Many of the large national laboratories are users, including Bell Canada, Oak Ridge, Pacific Northwest, Lawrence Livermore, Sandia, Brookhaven, National Renewable Energy Laboratory, SLAC, FermiLab, Lawrence Berkeley, NRC Canada, CERN, NIST, NASA, JPL, and NIH.

The most popular pages on the site recently have been Peak Finding and Measurement, Smoothing, Integration, Deconvolution, InteractivePeakFitter, and Signal Processing Tools. About 50% of the page views originate from search engines (80% of those using Google). The most common search keywords used are: "peak area", "convolution", "deconvolution", "peak detection", "signal processing pdf", "findpeaks matlab", "Fourier filter", and "smoothing". About 40% of the traffic comes from direct links (bookmarks or typed URLs) and about 10% comes from referring websites, usually from Wikipedia or from MathWorks. Unfortunately, page loads and search terms have become almost completely encrypted in recent years, so I can no longer tell which pages are being viewed and what is being download. (Interestingly, that is not the case with Interactive Computer Models for Analytical Chemistry Instruction, which has only 75% encryption).

There have been over 100,000 downloads of my software and documentation files, currently averaging several hundred file downloads per month, from both my web site and from my files on the Matlab File Exchange. The most commonly downloaded files are IntroToSignalProcessing.pdf, PeakFinder.zip, ipf12.zip, CurveFitter....xlsx, iSignal6.zip,  ipeak7.zip, PeakDetection.xlsx, and the complete site archive SPECTRUM.zip. 
 
What factors influence the number of page views from different countries?  The tools of data analysis, specifically regression via LINEST, can help answer this question. Obviously, one would expect that a country's population would be a factor, but it turns out that the correlation between log(page loads) and log(population) is very poor, with a coefficient of determination (log-log correlation coefficient or R2 value) of only 0.36 (n=163 countries; over 160,000 total page loads over the period from 2008 to 2017; graphic link). Note that because of the very large range of population sizes, I did a log-log correlation in order to prevent the results from being totally dominated by the top few countries.

I also investigated the effect of other factors that might be more specific to the language and subject matter of my particular site, including
All of that information is freely available on the internet (graphic link). By a good margin, the most influential factor was the research and development budget ,for which the R2 value was 0.76. This is perhaps not surprising given that my site concerns a very narrow and specialized topic: the technical aspects of computerized scientific data processing.

A log-log multilinear regression on all 5 of these factors together yielded a R2 value of 0.84 (n=53 countries for which all 5 factors were reported), which is a modest improvement over the research and development budget alone.

For an Excel spreadsheet with all these data and calculations (between 2008 and 2015), see FinalCountriesSummary.xlsx


This page is part of "A Pragmatic Introduction to Signal Processing", created and maintained by Prof. Tom O'Haver , Department of Chemistry and Biochemistry, The University of Maryland at College Park. Comments, suggestions and questions should be directed to Prof. O'Haver at toh@umd.edu. Site last updated July, 2022.