If we want to calculate chi square test for 3 by 2 or 4 by 2 or more than 2 by 2 consistency table, what is the procedure in SPSS?
If some cell has expected value less than 5 then what will be the p value as Fisher's Exact Test is not given for more than 2 by 2 consistency table?
I used crosstab option and used p value given in Asymp. Sig. (2-sided) column. Am I correct?
For tables larger than 2x2, the chi-square approximation can be good even if some expected counts are less than 5. The following rule of thumb (for tables larger than 2x2) is from a well-known textbook:
The chi-square approximation is good if "no more than 20% of the expected counts are less than 5 and all individual expected counts are 1 or greater" (Yates, Moore & McCabe, 1999, p. 734).
Re 2x2 tables with one or more expected counts < 5, the N-1 chi-square test is usually a much better choice than Fisher's exact test. See my notes at the link below, and Ian Campbell's website, which I link to in my notes.
https://sites.google.com/a/lakeheadu.ca/bweaver/Home/statistics/notes/chisqr_assumptions
I hope this helps.
You click Analyze -> descriptive statistics -> cross tabs. The you indicate your two columns. The click statistics and mark chi-square and click ok. if you want fishers you ask for an exact test also.
Chi square test and Fisher's exact test using SPSS can be done for 2*2/3*2/N*N tables using same procedure. The only difference is that both odd ratio and relative risk can be calculated only for 2*2 tables.
If i am using more than 2 by 2 consistency table in spss and if there is some cells that have expected frequency less than 5 did i need to calculate Fisher's Exact Test or I can use normal pearsons chi square value?
As I guess(not sure) fisher test is suitable for contingency 2x2 tables, not more than that and if calculated for more than 2 by 2 table using SPSS , calculation is often approximated using a Monte Carlo approach.
Yeah you still can use Monte Carlo's. Check the highlighted text in the attached file.
Fisher Exact test can be calculated for R*C cell. go to cross tab option then click to exact option then click to radio button of exact then continue. it will give you the fisher exact test value for R by C table
The chi-squared test for independence between two categorical variables
uses a normal approximation to generate a p-value. The approximation
is appropriate when the expected value under independence is large
enough in each cell (say at least 5). For 2 * 2 tables, this assumption can
be avoided by using Fisher’s exact test. However, for r * c tables, SPSS
does not supply an exact test by default. Exact tests can be generated for
any size cross-tabulation by permutation (e.g., for details, see Manly, 1999).
SPSS provides permutation tests for all test statistics used for two-way
classifications after clicking the Exact… button in the Crosstabs dialogue
box and checking Exact or Monte Carlo in the resulting sub-dialogue box.
The procedure is computationally intensive. SPSS sets a time limit and
offers to sample a set of permutations (Monte Carlo option). In the latter
case, the precision of the p-value is evaluated by a confidence interval (Landau, Sabine, and Brian Everitt. A handbook of statistical analyses using SPSS. Vol. 1. Boca Raton, FL: Chapman & Hall/CRC, 2004).
For tables larger than 2x2, the chi-square approximation can be good even if some expected counts are less than 5. The following rule of thumb (for tables larger than 2x2) is from a well-known textbook:
The chi-square approximation is good if "no more than 20% of the expected counts are less than 5 and all individual expected counts are 1 or greater" (Yates, Moore & McCabe, 1999, p. 734).
Re 2x2 tables with one or more expected counts < 5, the N-1 chi-square test is usually a much better choice than Fisher's exact test. See my notes at the link below, and Ian Campbell's website, which I link to in my notes.
https://sites.google.com/a/lakeheadu.ca/bweaver/Home/statistics/notes/chisqr_assumptions
I hope this helps.
A Fisher's exact test for more than 2*2 table.
https://www.jstor.org/stable/2288652?origin=crossref&seq=1#page_scan_tab_contents .
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Statistical Software
- SPSS