0
$\begingroup$

I have created a dataset starting from a series of multiple choice (3 choices) questions.

Example of question:

1) What color is Hulk?
    A) Green
    B) Red
    C) Pink

My dataset looks like this (I have computed many others statistics for each choice but the dataset down there is a good approximation of my real dataset for the sake of this question):

The dataset:

| question | frequency on Google | freq. on Google | freq. on Google | Correct |
|    id    |     of answer A     |  of answer B    |   of answer C   | answer  |
|––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––|
|    1     |        300          |       150       |       100       |    A    |
|––––––––––|–––––––––––––––––––––|–––––––––––––––––|–––––––––––––––––|–––––––––|
|    2     |         9           |       100       |       80        |    B    |
|––––––––––|–––––––––––––––––––––|–––––––––––––––––|–––––––––––––––––|–––––––––|
|    3     |       1000          |       400       |       800       |    A    |
|––––––––––|–––––––––––––––––––––|–––––––––––––––––|–––––––––––––––––|–––––––––|
|    4     |        35           |       50        |       125       |    C    |
|––––––––––|–––––––––––––––––––––|–––––––––––––––––|–––––––––––––––––|–––––––––|
|   ...    |        ...          |       ...       |       ...       |   ...   |
|––––––––––|–––––––––––––––––––––|–––––––––––––––––|–––––––––––––––––|–––––––––|

My goal is to find the correct answer given the variables I compute

Now here comes my question: Is the variable Correct answer really a categorical variable? I feel like it is not truly a categorical variable because correct answer of question 1 (which is A in the example dataset) has NOTHING TO DO with the correct answer of question 3) which is still choice A. Like, the two A's are not comparable because the questions are different!

Do I risk to create a completely useless model? Or could it work? Thanks!

$\endgroup$
1
  • 1
    $\begingroup$ As a side note, this research may be somewhat flawed or limited unless care is taken to properly define "correct answer." For example, the Hulk has changed colors depending on context and stories. See for example: io9.gizmodo.com/… $\endgroup$ Mar 23, 2019 at 19:53

1 Answer 1

0
$\begingroup$

Your "correct answer" variable is indeed categorical, more precisely it would be a binary variable (the answer if either correct (1) or wrong (0)) and you don't care about what the exact answer is. But it should also be possible to analyse another categorical variable that would be compatible with a multinomial logit regression: "Type of answer" (Green, red, etc.) - You don't care anymore about determining whether the answer is right/wrong, you simply analyse the probability of providing an answer.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.