b. Computed only for a 2x2 table and compare the significance (p-value) of Chi-Square statistic with α=0.05. In our case: 0.028<0.05 => Reject null hypothesis about absence of association and accept alternative hypothesis. Conclusion:age and brand preference are associated with each other.
MKF2121 2 of 2 Dr. Stanislav Stakhovych Note: If significance of Chi-Square statistic would be higher than α=0.05 (for instance, 0.3>0.05), then the null hypothesis would not be rejected and the conclusion would be: age and brand preference are not associated. If we do not reject null hypothesis the analysis stops. Step 3: Determine the strength of association If null hypothesis is rejected and we established that two variables are associated, the next step is to see how strong this association is. Symmetric MeasuresValue Approx. Sig. Nominal by Nominal Phi .159.028Cramer's V .159.028Contingency Coefficient .157.028N of Valid Cases 190The Phi coefficient is 0.159 which indicates poor associated between variables. Recall that this coefficient varies from 0 (no association) to 1 (perfect association). Step 4: Interpret the pattern of the relationship between variables Before interpreting the pattern it is very useful to compute the percentages in the direction of the independent variable, across the dependent variable. In our case the independent variable is Age and the dependent variable is Brand Preference. Brand Preference * Age Group CrosstabulationAge Group Total Less than 40 40 and more Brand Preference Pepsi Count 4528 73% within Age Group 45.9%30.4% 38.4%Coca-Cola Count 5364 117% within Age Group 54.1%69.6% 61.6%Total Count 9892 190% within Age Group 100.0%100.0% 100.0%Interpretation. Younger people (<40) prefer Pepsi and Coke almost equally (46% and 54%), however, older people (≥40) have strong preference towards Coke (70%) rather than Pepsi (30%).
- Two '16
- Chi-Square Test, Null hypothesis, Pearson's chi-square test, Fisher's exact test