turn on suggestions

Auto-suggest helps you quickly narrow down your search results by suggesting possible matches as you type.

Showing results for

- JMP User Community
- :
- Discussions
- :
- Discussions
- :
- ANOVA related question

Topic Options

- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Printer Friendly Page

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Permalink
- Email to a Friend
- Report Inappropriate Content

Dec 20, 2015 10:54 PM
(2757 views)

I have following data to implement analysis of variation.

1>. I follow the statistic requirements to check its pre-condition such as normanility, homogerous,

2>. Finding

a>. it is non normal distribution, what's more, their residuel also is obviously non normal

b>. they is not homogerous in deviation

I apply some methods to transfer data such as

root squre,

log,

and Fit model> factor profliling> Box-cox transformation, then save the best transformation, all of ways result and check their normality and homogerous,

the result also cannot meet the prequisite of ANOVA

in JMP, what I can do to conduct ANOVA reasonably?

treatment | Data |

A | 28 |

A | 22 |

A | 54 |

A | 19 |

A | 32 |

A | 31 |

B | 7 |

B | 11 |

B | 30 |

B | 6 |

B | 11 |

B | 13 |

C | 6 |

C | 9 |

C | 26 |

C | 7 |

C | 7 |

C | 11 |

D | 177 |

D | 151 |

D | 110 |

D | 117 |

D | 135 |

D | 138 |

E | 184 |

E | 146 |

E | 131 |

E | 110 |

E | 134 |

E | 141 |

- Tags:
- anova

3 REPLIES

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Permalink
- Email to a Friend
- Report Inappropriate Content

Dec 21, 2015 6:15 AM
(2558 views)

Step 0: Plot your data.

I think you may find that you can answer your questions about your treatments without ANOVA (or any formal statistical tests) from a graph. Judging normality of your responses is difficult with only 6 data points per group. You might also look carefully at Treatments A, B, and C as they each have an "unusual" data point. You might also read up on multiple comparison tests as alternatives to ANOVA (in general).

- Tags:
- anova

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Permalink
- Email to a Friend
- Report Inappropriate Content

Dec 21, 2015 12:06 PM
(2558 views)

I'll also add that before analysis it's helpful if you inform readers a bit about the practical problem you are trying to answer, and the hypothesis you are testing. One could presume the hypothesis you are testing is H0: mu(A)=mu(B)=mu(C)=mu(D)=mu(E)...but unless you tell us...well we're only guessing. Then we'd need to know the stated alpha risk too...before you do any data collection or analysis.

Once you get to analysis...then Karen's step 0 is a must. I shout from the rooftops each and every day, "The three rules for successful data analysis are: Rule 1. Plot the data; Rule 2. Plot the data; Rule 3. Plot the data.

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Permalink
- Email to a Friend
- Report Inappropriate Content

Dec 23, 2015 7:37 AM
(2558 views)

If you are concerned about the assumptions of oneway ANOVA, then try selecting one of the non-parametric tests such as Wilcoxon from the Oneway hotspot (red triangle) or alternatively try Partition from Analyze > Modeling, this will tell you which levels of your treatment variable result in different response means without making the assumptions of ANOVA.

- Tags:
- anova