Q: Preference choice data analysis

Q: Preference choice data analysis

Post by Tony » Fri, 18 Jul 2003 10:13:51



I'm trying to figure out how to analyze/the best way to analyze data from
the following type of question:

Which of the following 2 superheroes do you like better:  superman or
batman?

__ superman
__ batman
__ like both the same

My data looks something like this:

3 / 25% superman
4 / 33% batman
5 / 42% like both the same

The hypothesis I'd like to test is:

H0:  Both superman and batman are liked equally.

If reject H0:  Batman is liked more than superman.

My suspicion is that given n=12, I cannot reject H0.

What test/statistic do I need to test this hypothesis?

Tony

 
 
 

Q: Preference choice data analysis

Post by Rich Ulric » Sat, 19 Jul 2003 04:52:23




Quote:> I'm trying to figure out how to analyze/the best way to analyze data from
> the following type of question:

> Which of the following 2 superheroes do you like better:  superman or
> batman?

> __ superman
> __ batman
> __ like both the same

> My data looks something like this:

> 3 / 25% superman
> 4 / 33% batman
> 5 / 42% like both the same

> The hypothesis I'd like to test is:

> H0:  Both superman and batman are liked equally.

> If reject H0:  Batman is liked more than superman.

> My suspicion is that given n=12, I cannot reject H0.

Definitely,  n=12   hardly offers any test.

For n=120,  you might  do a  "test for proportions."
That might be a t-test  using the 4 separate numbers.
The 'tied'  sentiments do not contribute to the difference,
but they *do*  help make it logical to consider the proportions
as being (to some extent)  independent -- instead of being
100%  minus < the other>.

Actually, using the 2 numbers, the logical test would
be the comparison that the split was 50%  so it might
be done as a binomial test:  Is the parameter  0.50?

Why are you asking?  
After all is said, n=12  is too small
for any test be show a difference,
unless there was a clear preponderance.  
And then you  would probably have to be
concerned about 'randomness.'

--

http://www.pitt.edu/~wpilib/index.html
"Taxes are the price we pay for civilization."  Justice Holmes.

 
 
 

1. In search of table analysis expert and Stokes et al: Categorical data analysis

Hi,
I am in search of an expert for table analysis based on PROC FREQ,
especially for higher order tables. While writing a tabular overview about
tests for table analysis (just one page, you'll love it), I noticed a
somewhat confusing presentation Chapt. 4 (2 x r, s x 2 tables) in Stokes et
al., my favourite source. I contacted the authors, however SAS Institute
forwarded that it may take a few weeks for her to respond. Well, I cannot
wait that long, so I would like to get in touch with a table analysis expert
to make sure that this overview is free of flaws, and if interested, to
confirm or (better) to reject my considerations. Anybody interested in table
analysis?
Annette

2. SMS Server Change

3. SURVIVAL ANALYSIS SHORT COURSE | CATEGORICAL DATA ANALYSIS SHORT COURSE

4. Do I really need these?

5. Personal preferences in data management

6. Team Xbox: DEAD?

7. CPU choice and video card choice

8. InstallShield problems

9. Req;Business Analysis - Data Warehouse

10. Excel Data Analysis ToolPak

11. New statistics add-in for Microsoft Excel - a great improvement over the Data Analysis Toolpak!!

12. Help w/ data analysis in Excel 7.0

13. data visualization and analysis software?