What analysis test should be used?

What analysis test should be used?

1. Qualitative (categorical)

   • One categorical variable
     • one categorical variable from a single population.
     • Chi-square Goodness-of-fit (x^2)
       • comparing the frequencies you observe in certain categories to the frequencies you might expect to get in those categories by chance using a contingency table.
   • Two categorical variables
     • For example you are asking adults which fizzy drink they prefer: Pepsi or Sprite, and comparing the answers given by each gender.
     • Pearson's Chi-square test (x^2)
       • comparing the frequencies you observe in certain categories to the frequencies you might expect to get in those categories by chance using contingency table.
       • Equation: x^2= SUM OF[ (observed-model)/ model]

2. Quantitative (measurement)

   • Relationships
     • trying to fit a linear model to the data to outline correlation, i.e. y= bx+c, this model would explain current data and predict future patterns.
     • One predictor (measurement)
       • Continuous
         • Degree of relationship (Pearson correlation) r^2
           • essentially calculates the gradient of the 'line of best fit', so if r^2=1 or -1 it has a perfect positive/negative correlation. As r^2 gets closer to 0, the correlation becomes much less significant.
         • Form of relationship (Regression)
           • simple regression analysis
       • Ranks
         • Spearman's rs
           • A bivariate correlation coefficient that works on ranked data. *Ranking the data reduces the impact of outliers. *first rank the data, then apply Pearson's equation(r^2)...
           • EXAMPLE: want to assess how creativity compares to the position awarded in a storytelling competition. Positions recorded and results of a creative questionnaire. Although they use numbers the ranks are technically categories because has no numerical value although the ORDER does matter..
     • Multiple predictors (multiple regression)
       • Multiple Regression analysis
         • An outcome is predicted by a linear combination of two or more predictor variables. Outcome= Y, and predictors= X. Each predictor has a regression coefficient 'b' associated with it... Y= (b0 +b1X1 +b2X2 +......bnXn) + E
         • Predictor variables must be chosen based on a sound theoretical rationale- "do NOT just select all random predictors, bung them all into a regression analysis and hope for the best".
   • Differences
     • Two groups
       • Independent measures/ between-subjects design
         • so when the data is normally distributed- use independent t-test, if they are not normally distributed use Mann-Whitney. *so if you put the data in order of DV and assign a group- if normally distributed the two groups should have separated i.e. AAAAAABBABBB
         • Independent sample t-test
           • establishes whether two means collected from independent samples differ significantly.
         • Mann Whitney
           • also tests the difference between 2 samples but is non-parametric.
       • Repeated measures/ within-subjects design
         • both look at differences between 2 conditions that are experienced by 1 group. If data is based on a assumption of normal distribution of ranks use related sample t-test, or Wilcoxen if not.
         • Related sample t-test
         • Wilcoxen signed-rank test
           • *NOT to be confused with Wilcoxen's rank-sum test that is similar to independent samples t-test. *Non-parametric version of related-samples so still looks at comparing the scores of 1 group in both conditions, however the test does not make assumptions of normal distribution.
     • Multiple groups
       • Repeated measures/ within-subjects design
         • Repeated measures ANOVA
         • Friedman's ANOVA
       • Independent measures/ between-subjects design
         • One independent variable
           • e.g. 3 groups(influenced differently i.e. luxury/bargain) of wine experts test 3 wines and mean average for each pt across the 3 wines produced for each group. So comparing 3 groups with 1 iv (average score given to wine)
           • One-way ANOVA
           • Kruskal-Wallis
         • Multiple independent variables
           • Factorial ANOVA