All of the calculations for the equivalence test for 2x2 crossover design assume that the treatment order for sequence 1 is the reference treatment followed by the test treatment, and that the order for sequence 2 is the test treatment followed by the reference treatment.

Let *Y _{ijk}* be the response for participant

If the response for either period is missing for a participant, then the data for that participant is omitted from the calculations.

Let *d*_{1k} and *d*_{2k} be defined as follows:

Term | Description |
---|---|

Sample mean of d, _{1k}k = 1, ..., n_{1} | |

S_{1} | Sample standard deviation of d, _{1k}k = 1, ..., n_{1} |

Sample mean of d, _{2k}k = 1, ..., n_{2} | |

S_{2} | Sample standard deviation of d, _{2k}k = 1, ..., n_{2} |

Term | Description |
---|---|

ν | Degrees of freedom |

n_{1} | Number of participants in sequence 1 |

n_{2} | Number of participants in sequence 2 |

Let *k*_{1} be the lower limit that you specify and*k*_{2} be the upper limit that you specify. By default, the lower equivalence limit, *δ*_{1}, is given by:

and the upper equivalence limit, *δ*_{2}, is given by:

However, if you select the option to multiply your limits by the reference mean, , then the limits are given by:

is the average of the average responses for the two reference periods.