1. Home
2. AMDUNNETT procedure

# AMDUNNETT procedure

Forms Dunnett’s simultaneous confidence interval around a control (R.W. Payne).

### Options

`PRINT` = string token Controls printed output (`interval`); default `inte` Form of the alternative hypothesis (`twosided`, `greaterthan`, `lessthan`); default `twos` Probability level for the confidence interval; default 0.95, i.e. a 95% confidence interval Saves the lower confidence limit Saves the upper confidence limit Save structure to provide the means; default uses the save structure from the most recent `ANOVA`

### Parameters

`FACTOR` = factors Define the model term whose means are to be compared Scalar or single-valued text for each factor to identify which of the means of the term is the control; default uses the reference level of the `FACTOR`

### Description

`AMDUNNETT` is useful when you want to compare several treatments with a control treatment, and use a critical value that controls the chance that any one comparison may be found significant when there are no true differences. (It is designed thus to take account of the fact that you are making multiple comparisons with the control.)

The `FACTOR` parameter lists the factors that define the treatment term whose means are to be compared. The means are usually taken from the most recent analysis of variance (performed by `ANOVA`), but you can set the `SAVE` option to a save structure from another `ANOVA` if you want to examine means from an earlier analysis. The `CONTROL` parameter specifies a list of scalars to identify the levels of the factors that correspond to the control, or you can use a string (or single-valued text) to identify the level of any factor that has labels. If `CONTROL` is unset, `AMDUNNETT` uses the reference level of the `FACTOR`.

The `METHOD` option defines the type of interval that is formed. By default `AMDUNNETT` forms a two-sided interval. If you set `METHOD=lowerthan`, a lower confidence interval is formed to assess the one-sided test of the null hypothesis that the treatment means are not lower than the control mean. Alternatively, you can set `METHOD=greaterthan`, to obtain an upper confidence interval to assess the one-sided test of the null hypothesis that the treatment means are not greater than the mean of the control.

The probability for the confidence interval is specified by the `CIPROBABILITY` option; the default 0.95 gives a 95% interval. The lower and upper values of the interval can be saved (in scalars) using the `LOWER` and `UPPER` options, respectively. By default the interval is printed, but this can be suppressed by setting option `PRINT=*`.

Options: `PRINT`, `METHOD`, `CIPROBABILITY`, `LOWER`, `UPPER`, `SAVE`.

Parameters: `FACTOR`, `CONTROL`.

### Method

`AMDUNNETT` uses the methods of Dunnett (1955, 1989); also see Hsu (1996) Chapter 3.

### Action with `RESTRICT`

If the `Y` variate in the original `ANOVA` was restricted, only the units not excluded by the restriction will have been analysed.

Dunnett, C.W. (1955). A multiple comparison procedure for comparing several treatments with a control. Journal of the Americal Statistical Association, 50, 1096-1121.

Dunnett, C.W. (1989). Algorithm AS251 Multivariate normal probability intervals with product correlation structure. Applied Statistics, 38, 564-579.

Hsu, J.C. (1996). Multiple Comparisons Theory and Methods. Chapman & Hall, London.

Procedures: `AMCOMPARISON`, `AUMCOMPARISON`, `EDDUNNETT`, `CONFIDENCE`, `STEEL`, `VMCOMPARISON`.

Commands for: Analysis of variance.

### Example

```CAPTION  'AMDUNNETT example',\
!t('Data from Table 1 of Dunnett, C.W. (1964),',\
'New tables for multiple comparisons with a control.',\
'Biometrics, 20, 482-491'); STYLE=meta,plain
FACTOR   [NVALUES=80; LABELS=!t(A,B,C,D)] Treatment
FACTOR   [NVALUES=80; LEVELS=!(1,3,5,7)] Time
GENERATE Time,5,Treatment
VARIATE  [NVALUES=80] Fat