Hi Mr Cottrell, I'm a newly Gretl-addicted user. Since I discovered Gretl a few months ago, I gave up developing my rudimentary Visual Basic statistical software and geared up for C programming. Meanwhile, my point today is about Gretl's correlation matrix.
1. The "Correlation matrix " menu and the "corr" command don't yield the same output if less than 4 variables are selected. The later's output doesn't display the number of observations used neither the "5% critical value (two-tailed)". 2. While computing a correlation matrix (through the menu path or the command line) does the number of observations (n) and the 5% critical value that Gretl prints out really make any sense since Gretl's pairwise correlation depends on variables' missing values? Virtually, every pairwise correlation coefficient may have been drawn on different observations, with different 5% critical values. 3. From a regression analysis point of view I think that it would be more interesting to have, by default, a casewise matrix correlation instead of Gretl's pairwise matrix correlation, that uses the same observations throughout the variables selected by deleting all observations having at most one missing value. 4. One can always use the pairwise correlation (that can be an option to add) but if it's the case it should be noted the number of observations used for each correlation coefficient and the coresponding 5% critical value. Thank you for having offered us a very great Gretl! Arthur
