CALM

Extracting Hypotheses from Regression Codes

CALM can take existing coding patterns and decode them to determine the hypotheses being tested. The examples below use traditional and commonly-taught regression codes (dummy codes, effect codes, etc.) and extract the hypotheses implied by each code vector.

The package takes matrix input of the coding pattern (columns represent coded vectors, rows represent the groups being represented) and outputs the linear contrasts hypotheses being represented (the rows represent the weights for the contrast, the columns represent the groups being compared).


Dummy Coding

Specify a matrix of dummy codes and then use CALM to decode it into hypotheses.

ex_dummy <- matrix(c(
    1,1,0,
    1,0,1,
    1,0,0),
    ncol=3,
    byrow=TRUE)
calm.decode(ex_dummy)
##          GROUP 1 GROUP 2 GROUP 3
## HYPOTH 1       0       0       1
## HYPOTH 2       1       0      -1
## HYPOTH 3       0       1      -1

This shows that the intercept is equivalent to the mean of the last group. THe additional parameters represent the difference between each group and the last group.

Effect Coding

Specify a matrix of effect codes and then use CALM to decode it into hypotheses.

ex_effect <- matrix(c(
    1,1,0,
    1,0,1,
    1,-1,-1),
    ncol=3,
    byrow=TRUE)
calm.decode(ex_effect)
##          GROUP 1 GROUP 2 GROUP 3
## HYPOTH 1  0.3333  0.3333  0.3333
## HYPOTH 2  0.6667 -0.3333 -0.3333
## HYPOTH 3 -0.3333  0.6667 -0.3333

This shows that the intercept is equivalent to the mean of the group means. The additional parameters represent the difference between each group and the grand mean. The last group is not compared.

Repeated Coding

Specify a matrix of repeated codes and then use CALM to decode it into hypotheses.

ex_repeated <- matrix(c(
    1,0,0,
    1,1,0,
    1,1,1),
    ncol=3,
    byrow=TRUE)
calm.decode(ex_repeated)
##          GROUP 1 GROUP 2 GROUP 3
## HYPOTH 1       1       0       0
## HYPOTH 2      -1       1       0
## HYPOTH 3       0      -1       1

This shows that the intercept is equivalent to the mean of the first group. The additional parameters represent the difference between each group and the previous group.