Hello fellow scholars, I would like assistance with a Chi-square analysis dealing with two independent groups. In one research article I was reading, the researcher was studying US attitudes towards transgender individuals using a specific survey and how their results compare to responses from a different study using the same survey (this time in Sweden). They said that their response frequencies from this survey will be comparable to attitudes in Sweden- in that both surveys will be mostly positive. This was measured using a chi-squared comparing the frequency of responses between this study and those obtained for questions on the Landén and Innala’s (2000) Attitudes Toward Transgender National Survey. Link to study by Kooy (2010):
http://humboldt-dspace.calstate.edu/bitstream/handle/2148/693/Kooy%20Thesis%202010.pdf?sequence=5
The researcher was able to get this result: X² (2, n = 126) = 45.56, p < .001, φc = 0.43 for the following question: Transgender persons should have the opportunity to change their name
Numbers can be found for both studies when observing Table 4 in the results section of the linked study. I cannot understand how the researcher came to this conclusion? I've tried using an online Chi-square calculator and keep getting X²= 40.8049. Any help would be appreciated. Thank you.
FREQUENCY counts using chi square is not the same as X2 = (O - E)2 / E. This equation is used for quantitative value. For frequency analysis, use the 2x2 table. The equation for chi square in frequency analysis is:
X2 = A / B
A = (n - 1)(ad - bc)2
B = (a + b)(a + c)(b + d)(c + d)
The 2x2 table may be arranged as:
=========================
class 1 class2
sample 1 a b
sample 2 c d
=========================
Count the frequencies of a,b,c, and d.
REFERENCE: Gopal Kanji, 100 Statistical tests
https://www.pdfdrive.com/100-statistical-tests-e97177.html
Paul Louangrath Ok so I am assuming it was done incorrectly in the original study? I was using a 2x2 table before but just wasn't sure if that was correct or not. I am doing something very similar in my study as I will be comparing my frequencies with those from Kooy's (2010) study.
If one compares both studies, I get the same result as Alejandro:
Table of frequencies:
Response-----------Current--------------L&L2000
Yes-----------------------117-------------------427
No--------------------------3-----------------------94
Missing-------------------6----------------------147
X-squared = 41.172, df = 2, p-value = 1.147e-09
A test for only the current study would be a one sample chi-square with predefined expected frequencies.
Jos Feys First of all, thank you for the response! I'm glad that I was on the right track. Just to make sure i am understanding all of this, would this be a 2x2 design? It looks like the study was able to just compare the frequencies for "yes" responses with the L&L (2000) one.
When I look at the question and compare for just "yes" responses between both tables, my 2x2 looks like this:
--------------------------------------Yes------------------------Other (number of people that didn't endorse yes)
Kooy (n=126)----------------117------------------------ 9
L&L2000 (n= 668)---------428------------------------ 248
For comparison of just Yes responses, I get X²=40.8049, P < .01.
When the study does it however, they get: X² (2, n = 126) = 45.56, p < .001, φc = 0.43
I'm not 100% how they did this, I'll probably take it up with my supervisor but I wanted to try and figure it out on my own.
Jos Feys So I spoke with my supervisor, what she suggested was that it was a chi square goodness of fit analysis. The method you showed me though (chi square contingency), is correct and my supervisor said that it's a better method to use. Just wanted to say thank you again for your assistance, I'm back on track with my data analysis!
- Badges
- Science topic
- Similar topics
- Statistics
- Survey