Chi-Square OK for this Contingency Table?

Hi,

I'd like to check and see if I'm using the right stats in a
particular application, and I'd be most grateful for some advice.

I've got the following data grid:

Statement 1 Statement 2

Speaker A C1: % Correct C4: % Correct

Speaker B C2: % Correct C5: % Correct

Speaker C C3: % Correct C6: % Correct

Data was generated by an experimental survey where subjects were
randomly assigned across these six conditions. The experiment
presented each S with one statement. The statement came from either
Speaker A, B or C, and consisted of either Statement 1 or Statement 2.
The dependent variable is dichotomous: it measures whether each
individual subject succeeded or failed at grasping the point of the
statement that they were exposed to.

I am interested in determining whether the percentage/proportion of Ss
who successfully grasped the statements differs significantly across
conditions.

My approach so far has been to run chi-square analyses to determine
whether the the percentage of successful "graspers" (ouch)
differs across speakers. In other words, does the percent of Ss
correctly grasping Statement 1 differ across Speakers A, B and C?
(e.g, are the percentages in C1, C2 and C3 significantly different?)
Similarly, does the percentage of Ss correctly grasping Statement 2
differ across Speakers A, B and C? (do the percentages in C4, C5 and
C6 differ significantly?)

I would also like to determine if the percentage of Ss who are
"successful graspers" (ouch again) differs across the two Statements
for each Speaker. In other words, does C1 differ from C4? Deos C2
differ from C5? Does C3 differ from C6? I have been using Fisher's
exact test for these calculations.

I'm pretty confident that these tests are OK to use, but I'm worried
by the piecemeal quality of this approach. To make an analogy, it's
kind of like using a series of specific T-tests, when you really
should be doing an Anova as a single, overall test.

So, I guess the bottom line is two basic questions.

1) Is there an overall test for proportional/percentage data for a 2X3
table, or is my current chi-square/Fisher's approach OK?

2) If there is such an overall test, is there a way you can still
assess whether differences across two conditions are signficant? (C1
vs C2, for example, or C1 vs C4?

Whew..if you're still with me, I would *greatly* appreciate an
answer. Thanks much!

Political Psych Scholar

.

Relevant Pages

  • Re: Chi-Square OK for this Contingency Table?
    ... correctly grasping Statement 1 differ across Speakers A, ... does the percentage of Ss correctly grasping Statement 2 ... "successful graspers" differs across the two Statements ...
    (sci.stat.consult)
  • Re: Chi-Square OK for this Contingency Table?
    ... correctly grasping Statement 1 differ across Speakers A, ... does the percentage of Ss correctly grasping Statement 2 ... use this site for exact tests on up to 6x6 tables:http://www.physics.csbsju.edu/stats/exact_NROW_NCOLUMN_form.html ...
    (sci.stat.consult)
  • Re: Chi-Square OK for this Contingency Table?
    ... correctly grasping Statement 1 differ across Speakers A, ... "successful graspers" differs across the two Statements ... exact test for these calculations. ...
    (sci.stat.consult)