Can we rely to auto selecting statistic tests in SPSS2? For e.g. mann whitny for two group comparision or kruskal whallis test for more than two comparision in independent samples nonparametric tests.
You can rely on the program to automatically select which tests to use, but you're better off knowing which test to use and why. The output for the nonparametric tests will show you the null hypothesis (automatically generated), the test it used, significance level, and conclusion to reject or accept. This is nice and neat and usually makes sense, but it is far better to know your data and the question you want answered by the statistics. The nonparametric test menus can actually help guide you in this process. It's possible that the program will select the wrong test; also possible that it will perform some tests that are irrelevant.
You can rely on the program to automatically select which tests to use, but you're better off knowing which test to use and why. The output for the nonparametric tests will show you the null hypothesis (automatically generated), the test it used, significance level, and conclusion to reject or accept. This is nice and neat and usually makes sense, but it is far better to know your data and the question you want answered by the statistics. The nonparametric test menus can actually help guide you in this process. It's possible that the program will select the wrong test; also possible that it will perform some tests that are irrelevant.
I totally agree with Jerry Miller: it is better to know your data (check it using exploratory data analysis before run any statistical tests) and the question you want answered by the statistics (even better asking from the beginning of the experimental proposal, i.e, by experimental design). It will save headaches!