Dummy coding, AA.INP

[jbruggin]

Section 2.1, chapters 6 and 7 and the answers to various posts have convinced me that dummy variables (e.g., sex) need to be coded, for example, as '1 0' for males and '0 1' for females rather than 1 for males, 0 for females. Thus, I'm confused by the example on page 2-7 (from BLCKDUCK.INP) where age appears to be coded as 0 for subadult and 1 for adult. Is this correct?

I didn't get the file AA.INP referred to in Chapter 6 with my recent download of MARK. Is it still available?

[cooch]

Section 2.1, chapters 6 and 7 and the answers to various posts have convinced me that dummy variables (e.g., sex) need to be coded, for example, as '1 0' for males and '0 1' for females rather than 1 for males, 0 for females. Thus, I'm confused by the example on page 2-7 (from BLCKDUCK.INP) where age appears to be coded as 0 for subadult and 1 for adult. Is this correct?

You need to distinguish between dummy variable oding in the design matrix, and the frequency coding in the input file. The former is discussed at length in Chapter 7, and the latter is discussed at length in Chapter 2.

I didn't get the file AA.INP referred to in Chapter 6 with my recent download of MARK. Is it still available?

The example data files referred to in 'the book' are not distributed with MARK, but are available on the website where you downloaded 'the book' (last item in the drop-down list).

[jbruggin]

OK, my question referred primarily to chapter 2 of ‘the book’ as I am primarily concerned with input file formatting at this point. Sorry if I’m being obtuse but I’ve reread it (again), and remain confused. Near the top of page 2-4 is the following example of how to code individual encounter histories:

110000101 1 0;
110000101 1 0;
110000101 1 0;
110000101 1 0;
110000101 0 1;
110000101 0 1;

Where “the coding ‘1 0’ indicates that the individual is a male, and ‘0 1’ indicates the individual is a female.”

In contrast, later in chapter 2 (page 2-7) this example of covariate coding is given:

/* 01 */ 1100000000000000 1 1 1.16 27.7 4.19;
/* 04 */ 1011000000000000 1 0 1.16 26.4 4.39;
/* 05 */ 1011000000000000 1 1 1.08 26.7 4.04;
/* 06 */ 1010000000000000 1 0 1.12 26.2 4.27;
/* 07 */ 1010000000000000 1 1 1.14 27.7 4.11;
/* 08 */ 1010110000000000 1 1 1.20 28.3 4.24;
/* 09 */ 1010000000000000 1 1 1.10 26.4 4.17;

Here the first covariate “is a dummy variable representing age (0 = subadult, 1 = adult)” in a single column.

Are not both examples dealing with frequency coding? If I need 3 columns to code for sex in the first example, why don’t I need 3 to code for age in the second? Do I need to digest chapter 7 in order to understand this aspect of chapter 2?

Thanks.

John

[ganghis]

John,

In the first case, sex is treated as a group, and in the second it is treated as a covariate. Which one you want to use will depend on your purposes... you generally will need to do more work in the design matrix if you treat it as a covariate. Note if there were more than 2 groups (e.g. male, female, and unknown) you would almost always want to code them as different groups in the input file.

-Paul

[cooch]

Are not both examples dealing with frequency coding? If I need 3 columns to code for sex in the first example, why don’t I need 3 to code for age in the second? Do I need to digest chapter 7 in order to understand this aspect of chapter 2?

Thanks.

John

In INP file, you're not 'coding' for anything - each group (or sub-group...) needs a column. I suspect you're confusing design matrix coding with basically having a column of frequencies for each group in the INP file.

So, if your groups are 'males' and 'females' (as per the example), you have two columns of frequencies - one for males, and one for females. If you see

Code: Select all

1011101  1 0;


the last two columns are not coding columns, they are frequency columns.

In the second example, with individual covariates, 'sex' is itself coded as an individual covariate. And, it is indeed coded, so only one column is needed.

Paul has explained why you might choose one over another.