Continuous or Discrete: Likert-type Time

#1
Hi!

Sorry for asking a very basic question. I'm not good at stats and can't reach someone who knows. I searched the net with many different keywords but failed to find any answers.

So, Are these variables Discrete or Continuous?

1. Approximately how long have you had your MySpace account?
a.) 6 months
b.) 1 year
c.) 1.5 years
d.) 2 years
e.) 2.5 years
f.) 3+ years

2. Approximately how many minutes per day do you spend on MySpace?
a.) 10 or less
b.) 10–30
c.) 31–60
d.) 1–2 h
e.) 2–3 h
f.) 3+ h

I appreciate any help :confused: :rolleyes: and Thank you in advance...
 
#3
Is this homework and are discrete/continuous the only options?
Thanks for responding Dason!

This is part of my thesis. I am planning to conduct a regression analysis. If the variables are continuous I have more options. Ordinal and Discrete ones limit the study.
 

Dason

Ambassador to the humans
#4
Well question 1 has a discrete response but question 2 I would classify as ordinal.

What is the overall goal? Do you want to use these as predictors or as responses?
 
#5
Well question 1 has a discrete response but question 2 I would classify as ordinal.

What is the overall goal? Do you want to use these as predictors or as responses?
I want to use them as responses. I have a personality scale as predictor. I want to conduct a regression analysis as: (IV) personality -> demographics (DV)

Would it be "completely wrong" to use them as continuous dependent variables in a multiple regression?
 

noetsi

No cake for spunky
#6
It varies but commonly if you have 5 or more categories likert scale data is interval enough to use in OLS. The key issue is less if its discrete than if its multivariate normal (I had a fun discussion on this topic with a professor in this area today - he noted that contrary to the common wisdom in stats books, even binary dependent data could often be used in OLS as a dependent variable).
 
#7
It varies but commonly if you have 5 or more categories likert scale data is interval enough to use in OLS. The key issue is less if its discrete than if its multivariate normal (I had a fun discussion on this topic with a professor in this area today - he noted that contrary to the common wisdom in stats books, even binary dependent data could often be used in OLS as a dependent variable).
Thank you for responding noetsi.

So, do you say that I can use those two variables as -continuous- dependent variables in regresion?
 

noetsi

No cake for spunky
#8
Well it depends on many things. First, does you audience accept it is valid. The only way to know that is to ask them. Second, how exact do your answers have to be. If you use non-interval data in OLS it will degrade the results. Commonly it won't degrade the results enough to matter, but you have to decide if what is lost in the process is worth it. At the least you should run the data in both OLS and say ordered logistic regression and see how much the results vary. Even if likert scale data works generally for OLS does not mean it will in your case - you have to look at the results to determine this.

You should also look at how skewed the data is (any commerical software does this test). If its beyond 3 you might want to consider transforming it to make it work better in OLS.
 
#9
Then, I will try them as both continuous and discrete. Check if there is any significant difference in the results. And use continuous one if it seems OK. So, this is normal regression.

If the results look strange, then I will treat both of them as discrete. So, this is logistic regression.

Is that all right?

ps: by the way; thank you both for helping me.