[Q] How to analyze a variable that can be both quantitative and qualitative?

[P300]

I have the independent variable "music tempo" that can take multiple quantitative values (e.g., 90 BPM, 100 BPM, and 110 BPM). My dependent variable is "length of the music" (needed to feel absorbed by it) in seconds. However, I also gave the option to my participants to say that they needed "MORE" time to feel absorbed or to say that they would "NEVER" be absorbed by this type of music. Therefore, the dependent variable "length of the music" can take quantitative and qualitative values. Does anyone know how to analyse this variable? How can i make an ANOVA to see whether there are differences in the length of the music depending on tempo with this type of dependent variable?

I think an option would be to analyse the quantitative and qualitative answers of the dependent variable separately. However, if I do this i lose a lot of data, because some participants answered with a time in seconds for one tempo, but with "MORE" or "NEVER" in other tempos.

Thank you in advance!

 

[Karabiner]

So you can rank participants according to lenght, the second highest rank being "more" and the highest rank being "never" (=unlimted time).
Or, you treat "never" as "not applicable" and analyse these cases seperately; it depends on the context and the background of your research.

Then you can for example correlate tempo with time using the Spearman rank correlation coefficient rho.

With kind regards

Karabiner

 

[gw1]

ANOVA is not used to test discrete (non-continuous ('i.e. Never, More time)) dependent variables. A generalised linear model (e.g. binomial / multinomial) may be worth considering for discrete dependent variables. I'm not familiar with a model that will output both continuous and discrete dependent probabilities at the same time.

If the qualitative dependent variable enables subjects to answer 'more time' or 'never', does the test have a maximum duration for playback? Could another option be to have 'seconds' as a second independent variable with one or more levels (durations of playback), and one dependent variable with e.g. 'really feeling the groove', 'more time', 'never'? In such a case, could a multinomial model (if you have a predetermined sample size) be useful for analysis? See multinomial response data page 628 on the link https://pdfroom.com/books/an-r-companion-to-applied-regression/andLV7l42e3/amp (page 309 in print). Consider assumptions for the model.
If the discrete dependent variable only had two options, a binomial model could be considered.

If you're testing one particular track for the minimum individual's time needed to feel the groove, 'seconds' as a continuous dependent variable seems better to me. Does the timer have milliseconds? Sokal and Rohlf (Biometry 3rd ed. page 14) recommend having a minimum 30 unit steps (maximum 300) between the minimum and maximum recorded values used in a data set for statistical inference.