Which data types & correlation method?

nini

New Member
#1
Dear all,


I'm a new graduate student and not yet proficient at statistics in research. Please help me confirm/correct the following questions.

* I want to examine if streaming video viewing styles correlate with reading frequency, comprehension, and speed.

1. one kind of data is 4 types of video viewing styles --> When I enter data in SPSS, do I specify the data type as "nominal"? If not, please correct me.

2. other kinds of data are reading freqeuncy, comprehension, speed collected from likert scale --> Are these specified as "ordinal data" in SPSS? If not, please correct me.

3. Which kind of correlation method should I use?

4. To test the significance of the results, which test should I use?


I'd appreciate any help! Thanks!


nini
 

noetsi

No cake for spunky
#2
1 It depens if the types can be logically ordered or not. If they can as in Most satisfied, less satisfied, not satisfied, then your data is ordinal. If there is an ordering in your streaming (say the first level is slowest and the last level if highest in speed) and each level is distinct from the other then you should specify the categories as ordinal. Otherwise they are nominal.

2) If they are likert scale then you should specify them as ordinal (although if they have at least five categories, you have enough cases, and there is not serious non-normality you can use it as interval data in practice).

3) Spearman's Rho is suitable for ordinal data. If one of your variables are nominal then you either use chi square (which shows if variables are related but won't give you a correlation which can't be generated with nominal data) or you look at contigenncy tables. Alternately if you think one or more of the variables explains another you can use regression or ANOVA.

4) Chi square if they are nominal. You have lots of choices if they are ordinal. Regression (t test) and Gama are two examples.
 

nini

New Member
#3
Yes, if:

1. The data collected for reading frequency, comprehension, speed are specified as "ordinal" type of data (due to they are collected from likert scale)

2. Four video viewing styles are specified as "nominal" type of data


then, which statistic method can be used to examine the correlations between "nominal data" and "ordinal data" respectively or can be used to examine their correlations all at one time?


nini
 

noetsi

No cake for spunky
#4
As far as I know there is no corrlation for nominal variables although various forms of regression generate something similar. Neither Pearson nor Spearman's Rho (the common correlation statistics) is supposed to used for them.