help

help >

**RE: 2x3 Mixed Design ANOVA**Oct 11, 2018 06:10 PM | Andrew Zalesky

RE: 2x3 Mixed Design ANOVA

Hi Ana,

A simple design to test the ordering of time1 > time2 > time3 is to use a design matrix with two columns: Column 1 is filled with ones; Column 2 contains a 1 for time point 1, 2 for time point 2 and 3 for time point 3. Additional columns are needed to model the within-subject (repeated measures) effects. These would be the same the previous design. The contrast for a t-test would be [0 1 0 0 0...] and [0 -1 0 0 0...]. The first contrast tests t1 > t2 > t3, while the second contrast tests t1 < t2 < t3.

Note that this is a very simple design and assumes a linear increases (or decrease) over time, as well as other assumptions. There are more advanced models available that you might want to investigate.

If the gap between t1 and t2 is smaller than the gap between t2 and t3, you can put a value higher than 3 in the design matrix for time point 3.

Andrew

A simple design to test the ordering of time1 > time2 > time3 is to use a design matrix with two columns: Column 1 is filled with ones; Column 2 contains a 1 for time point 1, 2 for time point 2 and 3 for time point 3. Additional columns are needed to model the within-subject (repeated measures) effects. These would be the same the previous design. The contrast for a t-test would be [0 1 0 0 0...] and [0 -1 0 0 0...]. The first contrast tests t1 > t2 > t3, while the second contrast tests t1 < t2 < t3.

Note that this is a very simple design and assumes a linear increases (or decrease) over time, as well as other assumptions. There are more advanced models available that you might want to investigate.

If the gap between t1 and t2 is smaller than the gap between t2 and t3, you can put a value higher than 3 in the design matrix for time point 3.

Andrew

*Originally posted by Ana Coelho:*Thank you very much for your quick reply.

It's true that I want to test specific hypothesis. This is an aging study and I am using diffusion data to build structural connectivity. So, I want to test if structural connectivity will decrease with time but increases could also happen, so I want to test both hypothesis: time1 > time2 > time3; time1< time2 < time3. How can I do this?

Another thing, the time between acquisitions in the 3 timepoints is not the same for all subjects. From time1 to time2 it varies between 16 to 31 months and from time 2 to time3 between 25 to 55 months. Should I include this in the model as cofounds? If true, how can I do this?

Thanks in advance!

Best regards,

Ana

It's true that I want to test specific hypothesis. This is an aging study and I am using diffusion data to build structural connectivity. So, I want to test if structural connectivity will decrease with time but increases could also happen, so I want to test both hypothesis: time1 > time2 > time3; time1< time2 < time3. How can I do this?

Another thing, the time between acquisitions in the 3 timepoints is not the same for all subjects. From time1 to time2 it varies between 16 to 31 months and from time 2 to time3 between 25 to 55 months. Should I include this in the model as cofounds? If true, how can I do this?

Thanks in advance!

Best regards,

Ana

## Threaded View

Title | Author | Date |
---|---|---|

Ana Coelho |
Oct 10, 2018 | |

Andrew Zalesky |
Oct 10, 2018 | |

Ana Coelho |
Oct 11, 2018 | |

Andrew Zalesky |
Oct 11, 2018 | |

Ana Coelho |
Oct 12, 2018 | |

Andrew Zalesky |
Oct 15, 2018 | |

Ana Coelho |
Oct 15, 2018 | |

Andrew Zalesky |
Oct 15, 2018 | |

Ana Coelho |
Oct 18, 2018 | |

Andrew Zalesky |
Oct 18, 2018 | |

Ana Coelho |
Oct 19, 2018 | |