DYNAMISM(E) - Biannual Student Magazine June-2017 | Page 11

Most commonly in business research, two types of hypotheses are tested. • Relational Hypothesis- Examine how chang- es of one variable vary with changes in other. E.g. Advertisement increases sales. • Hypothesis about differences- examine how some variable varies from one group to an- other. E.g. Stress level varies among working women of IT and Non-IT sector. For univariate analysis, one more type of hy- pothesis is tested- hypothesis about differenc- es from some standard. E.g. More than 50% of males in Bangalore smoke. • Univariate test for metric data can be done through t test or z-test depends on sample size and whether standard deviation of the population is known or can be estimated from sample. • Univariate test for non-metric data can be done through chi square, test, K-S or Runs test to drawn inference about a population depending on the nature of the sample. Case 11 Bi-variate Analysis- Different types of statis- tical techniques are used to analyse bi variate data, based on- a) scale of measurement b) type of distribution c) purpose of the test (sig- nificant relation or significant differences). Nom- inal and Ordinal scales are forming categorical scale whereas Interval and Ratio scale comprise Continuous scale. Categorical variables follows non-normal distribution (Poisson, Binomial etc.) and continuous variables may follow a normal or non-normal (Poisson, Binomial etc.) distribution. To check whether the variable follows a normal or non-normal distribution, researcher can use K-S test or Shapiro Wilk Test. Box plot of all the observation of the variables also gives an idea about the distribution. Scale of measurement and type of distribution are the two factors to be considered to decide whether to use parametric or non-parametric test. Please refer the diagram below for Bi-vari- ate data analysis: Figure: 1 X Variable Y Variable Distri- bu-tion of Y variable Purpose Type of test Statistical Technique 1 Cat Cat Non Nor- mal Sig. Differ- ence Non Parametric test Chi square or Z test 2 Cont. Cont. Normal/ Non-Nor- mal Sig. Relation Parametric test/ Non-Parametric test Pearson’s Correlation, bi-variate regression/ Spearman’s Rank Corre- lation 3 Cat (X variable has 2 cate- gories) Cont. Normal/ Non-Nor- mal Sig. Differ- ence Parametric test/ Non-Parametric test T test (paired t test or independent t test)/ Mann Whitney test 4 Cat (X variable has 3 or more categories) Cont. Normal/ Non-Nor- mal Sig. Differ- ence Parametric test/ Non-Parametric test ANOVA test/ Kruskal Wallis test DYNAMISM(E)