4
$\begingroup$

I have data similar to that (fake values):

| Individual | Profession  | Software_A | Software_B |
|------------|-------------|------------|------------|
| A          | Teacher     |        4.5 |        3.2 |
| B          | Cop         |        5.0 |        2.0 |
| C          | Teacher     |        2.0 |        1.2 |
| D          | Cop         |        5.0 |        2.0 |
| E          | Hairstylist |        1.0 |        0.5 |

Need to test if mean rates of Software_A is greater than Software_B for some Professions.

I can conduct a Paired Sample T-test comparing means, but that would give me the overall difference, not within groups.

Is there something like a Paired Sample ANOVA?

Possible solutions

  1. Should I simply compute a new variable called Rate_Difference = Software_A - Software_B and conduct ANOVA with Profession?
| Individual | Profession  | Software_A | Software_B | Rate_Difference |
|------------|-------------|------------|------------|-----------------|
| A          | Teacher     |        4.5 |        3.2 |             1.3 |
| B          | Cop         |        5.0 |        2.0 |               3 |
| C          | Teacher     |        2.0 |        1.2 |             0.8 |
| D          | Cop         |        5.0 |        2.0 |             0.5 |
| E          | Hairstylist |        1.0 |        0.5 |             0.5 |
  1. Should I restructure data turning Software_A and Software_B into different cases and perform a Bivariate ANOVA on Rate versus Software versus Profession?
| Individual | Profession  | Software | Rate |
|------------|-------------|----------|------|
| A          | Teacher     | A        |  4.5 |
| B          | Cop         | A        |  5.0 |
| C          | Teacher     | A        |  2.0 |
| D          | Cop         | A        |  5.0 |
| E          | Hairstylist | A        |  1.0 |
| A          | Teacher     | B        |  3.2 |
| B          | Cop         | B        |  2.0 |
| C          | Teacher     | B        |  1.2 |
| D          | Cop         | B        |  2.0 |
| E          | Hairstylist | B        |  0.5 |
$\endgroup$
2
$\begingroup$

Unless there's something that I'm missing here, what you're looking to do is just to carry out a two-way repeated measures ANOVA. This will allow you to test whether software interacts with profession (i.e. does the effect of software vary from one profession to another).

You would need to ensure that each profession occurs a reasonable number of times in your data though. Maybe >5.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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