Is predictorMatrix doing right?

[iago-pssjd]

I'm not sure how to understand the use of predictorMatrix. After reading in https://stefvanbuuren.name/fimd/sec-modelform.html that in

> set.seed(1); imp <- mice(nhanes, print = FALSE)
> imp$predictorMatrix
    age bmi hyp chl
age   0   1   1   1
bmi   1   0   1   1
hyp   1   1   0   1
chl   1   1   1   0

The rows correspond to incomplete target variables, in the sequence as they appear in the data. A value of 1 indicates that the column variable is a predictor to impute the target (row) variable, and a 0 means that it is not used. Thus, in the above example, bmi is predicted from age, hyp and chl. Note that the diagonal is 0 since a variable cannot predict itself. Since age contains no missing data, mice() silently sets all values in the row to 0. The default setting of the predictorMatrix specifies that every variable predicts all others.

Then I would understand that:

1. first, as for age, I would understand that if I set all values of bmi row to 0, then bmi missings wouldn't be predicted. But it isn't the case.
2. further, if those row values only affect to the prediction of bmi, shouldn't the predictions of the other variables be equal (under set.seed(1))? (unless predictions for the other variables depend on the predictions made for bmi?)

Can you help me?

Thank you!

[stefvanbuuren]

Thanks for your interest.

I am afraid I do not understand what you do not understand. Perhaps running some examples might help to increase your level of comfort. If that results in questions that I can understand, I'd be perfectly willing to answer.

Hope this helps, Stef.

[stefvanbuuren]

Note: "mice() silently sets all values in the row to 0" is old behaviour, and not true anymore.