Help Needed with Encoding Knowledge Score Variable *

[Alex John]

Hello,

I am new here and encountered an issue while analyzing and running regressions for the knowledge score variable. Unfortunately, I don’t have access to the do-file that the previous biostatistician used to encode the `score_knowledge` variable.

The author has added results from additional knowledge questions and requested that I calculate and integrate these into the existing `score_knowledge` variable.

Here’s what I know:

1. **Example of an individual knowledge question**
- `k10_correct` is a recoded variable from `k10_expdate_12m` (Knowledge Q10).
- It’s a numeric variable (byte) with values:
- `0 = Wrong`
- `1 = Correct`

2. **Codebook of the `score_knowledge` variable**
- Type: Numeric (float)
- Range: 0–90 (in units of 10)
- Mean: 45.7251, Std. Dev: 16.6574

My Question:
Is it correct to calculate the updated knowledge score by:
1. Adding the total number of correct answers across all knowledge questions (existing and new),
2. Then dividing by the total number of knowledge questions to get the average score?

Alternatively, is it possible to reconstruct or modify the original encoding logic for the `score_knowledge` variable to ensure consistency?

Thank you for your guidance!

[George Ford]

what did the previous biostat say was done?

[Clyde Schechter]

Clearly, it would be best if you knew how the original knowledge score was calculated. Have you tried to contact the original biostatistician? If not, I would do that first. It would surprise me to learn that the biostatistician no longer has a record of the code if the study is sufficiently recent that one of the authors is still developing the knowledge test.

That said, range 0-90 in units of 10 suggests to me that there were 9 items in the scale and the score is 10 * the number of correct answers. If that is the case, you can just

gen new_score = old_score + 10*(number_additional_items_correct)

Of course, it is possible that some more complex scoring system was used, typically involving some weighting of the items, but in short scales like this, that would be uncommon.

Added: Crossed with #2.

[Alex John]

Thank you for your detailed answer. I contacted the PI and he helped me with the old items and their weights. Your help is appreciated