Outlier tests

Having outlier tests, such as the ones by Grubbs, Nixon or Nalimov, would IMO be a great feature for statistiXL. Surprisingly, only few statistical packages have those even though these tests are pretty simple as far as their algorithms are concerned.

Comments

  • I support this suggestion by Excelian, although my own preference would be for identification by standard deviations (with, of course, the ability to set the number of standard deviations that one wishes to use). A related 'wish' is for identification of missing data. Sure, we can find missing data by means of Excel's filter, but the kind of feature which would be helpful is a provision to compute ways of dealing with the missing data; for example, by inserting the variable mean or by imputing a value by multiple regression or, perhaps, the EM methodology. A spreadsheet is useful when cleaning data, and adding the suggested features would be a nice extra. Another nice extra not provided by Excel would be a provision for plotting error bars. Regardless of whether these suggestions are accepted, statistiXL is a great program, and I value what it offers already. smile.gif Stan
Sign In or Register to comment.