Hi there! I was just wondering if anyone would be able to share some insight on perfo= rming a one-way between-groups ANOVA (or a UNIANOVA) for three distinct gro= ups. One group had high-intensity math assessments, another had minimal math ass= essment, and the last group has had no math assessment. These are all pre-d= etermined data, and was therefore not subject to a traditional research des= ign where students are randomly assigned to each group. I would like to compare each groups' performance in a math class, and am wo= ndering if an ANOVA is appropriate, given how the assumption of the data be= ing randomly sampled is being violated. Any help would be appreciated. Thanks!

0 |

11/1/2016 10:52:58 PM

On Tue, 1 Nov 2016 15:52:58 -0700 (PDT), darko.giacomini@gmail.com wrote: >Hi there! > >I was just wondering if anyone would be able to share some insight on performing a one-way between-groups ANOVA (or a UNIANOVA) for three distinct groups. > >One group had high-intensity math assessments, another had minimal math assessment, and the last group has had no math assessment. These are all pre-determined data, and was therefore not subject to a traditional research design where students are randomly assigned to each group. > >I would like to compare each groups' performance in a math class, and am wondering if an ANOVA is appropriate, given how the assumption of the data being randomly sampled is being violated. > >Any help would be appreciated. > You can always perform the test, whatever test you wonder about. Is it convincing to anyone? It may or may not give you a number that is useful, in some sense, for comparing to some other number. For instance, if your computation shows that the groups are apparently equal, then lack of randomization would have to be /offsetting/ some other difference, instead of creating one -- which might or might not be conceivable. What statistical theory says is something like, "you cannot assert strong inferences-based-on-randomization when there is no randomization." Whatever inference you draw should have to explicitly take into account and try to account for the absence of randomization. This is why simple "observational studies" take second place to "randomized studies." -- Rich Ulrich

0 |

11/2/2016 6:10:32 AM