My sample is 131 television ads. Chi-square of ad categories gives significant result with each of 6 advertising appeals but the assumption of expected count less than 25% cells is violated. When i double my dataset, it solves the problem. Ethically can i do this? some othe way? Please someone urgently help.
There are several things that you could do. The simplest is to try to combine several categories in your contingency table. Another thing you could do is use an exact test (e.g. Fisher's exact test), but these are complex for tables larger than 2x2 and involve resampling procedures (randomization tests). A simpler alternative may be to use a log-likelihood test (a G-test), which is supposed to be less affected by small frequencies than chi-square tests. Any introductory statistics textbook (e.g. Zar) will tell you how to do a G-test.
