| Substitutes for the t-test |
|
|
|
|
|
|
|
|
|
|
|
|
| Replication is necessary for estimating the SD in the |
|||||||||||
| treatment and control groups, needed for the |
|||||||||||
| denominator of t. In the
absence of replication, a |
|||||||||||
| parametric model for s could be used, provided |
|||||||||||
| that the amount of variability can be accurately |
|||||||||||
| predicted from a single slide (or a single-slide, |
|||||||||||
| with
color reversal) experiment. |
|||||||||||