Agena

4.11.6 Houma, April 23, 2025 - Instead of a procedure you can now also pass a string as the `ignore' option to `utils.readcsv`: If given a string, then all lines in a file starting with this string will be ignored and skipped. For example, instead of > utils.readcsv('sunspots.csv', ignore = << x -> x[1] = '#' >>) to skip all lines starting with a hash, you can now write: > utils.readcsv('sunspots.csv', ignore = '#') - New `stats.card` returns the number of samples of any distribution, be it a table, sequence, register or double numeric array. You can pass a Boolean function that determines which samples to count. Also, by giving the `meanvar=true' option, the function - besides the count - also returns the arithmetic mean and variance, all three in just one pass over the structure examined, saving runtime. You may also prompt the function to either return the population or sample variance, and to alternatively use a special algorithm to avoid numeric overflow with very large values, see the Primer & Reference for further information on that. > stats.card([10, 20, 30, 40, 50]): 5 > stats.card([10, 20, 30, 40, 50], meanvar=true): 5 30 200 > stats.card([10, 20, 30, 40, 50], << x -> x > 20 >>): 3 - `stats.meanvar` can now be passed the following options: - When given the `optimised=true' option, the function more strictly tries to avoid numeric overflow with very large samples. The default is `false`. - The `sample=true' option forces the function to return the sample variance instead of the population variance. The default is `false`. You now can also pass a multivariate Boolean selection function and put its second, third, etc. argument right after the distribution. - New `stats.meanqmdev` returns the arithmetic mean and the quadratic mean deviation of any distribution. Mathemetically, the quadratic mean deviation is the population variance multiplied by the number of samples in a distribution. The function by default prevents overflow with very large samples. - `stats.skewness` and `stats.kurtosis`: you can now compute the sample skewness and sample kurtosis of a distribution by passing any second non-null argument. - The following new correlation functions have been added to the `stats` package: stats.besselj Bessel J correlation function stats.besselk Bessel K correlation function stats.circular circular correlation function stats.constant constant correlation function stats.cubic cubic correlation function stats.dampedcos damped cosine correlation function stats.dampedsin damped sine correlation function stats.exponential exponential correlation function stats.gaussian Gaussian correlation function stats.hole hole correlation function stats.linear linear correlation function stats.matern Matern correlation function stats.penta pentaspherical correlation function stats.power power correlation function stats.ratquad rational quadratic correlation function stats.spherical spherical correlation function stats.white white noise correlation function - New `cordic.cpow(a, b)` computes a^b using CORDIC arithmetic. `cordic` function names have finally been added to the scheme files, too. - Added files `data/airline.csv` and `data/sunspots.csv` to the installers, to play around with real-world time series with the `stats` package. See the respective *.txt files for credits and explanation.