# Biostatistics

TestCriteria

Chi-Square

Forindependent variable, chi-square tests if the paired observationsmade on two variables in a contingency table are independent. This isdone through calculation of test statistic, determining the degree offreedom and then comparing the test statistic with chi-squaredistribution (Wilson &amp Lipsey, 2001). In testing the dependentvariable, a relationship is created with the dependent variablethrough comparison.

Analysisof Variance

Todo ANOVA, t-test is carried out with one independent variable butwith two or more levels. In calculating t-test, the independentvariable is compared to a single dependent variable.

T-test

Theindependent sample t-test makes a comparison of the mean values of acharacteristic measured on a continuous scale between twomicro-groups that make a categorical variable. T-test of independenceevaluates if paired observations that are made on two variables in acontingency table are independent of each other (Wilson &amp Lipsey,2001).

WilcoxonRanked Sum Test

Thistest treats data on dependent variable as ordinal. For the test ofindependent variables, two random variables, x1 and x2 are used andthus null hypothesis are tested such that X1 X2. The variables x1 andx2 are from samples n1 and n2. Upon testing, there is use of averageranks.

Kruskal-WallisTest

Thetest is carried out in the same manner as ANOVA. For the dependentvariable, the test relies on ranks or ordinal data and the test iscarried out on two or more independent variables.

McNemar’sTest

Whentesting, researchers show dependence through alternative hypothesisand refute independence of the variables through null hypothesis.

LinearRegression

Linearregression establishes the relationship between the two variables,dependent and idependent variables. The regression can be simple ormultiple.

LogisticRegression

Forindependent variable, this regression is employed when the variableis categorical while the dependent variable has to be dichotomous.

Pearson’sCorrelation

Pearson’sCorrelation is used for nominal or ordinal independent and dependentvariables. Variables can have any number of distinct levels. It testswhether the distribution of the dependent variable is the same foreach level of the independent variables. Thus, the null hypothesis isthat there is no association between the variables.

Spearman’sCorrelation

Intesting the correlation, if the variables are independent then thecorrelation coefficient is 0 however the case is not the same forindependent variable given that the correlation coefficients onlyshow linear dependencies between any two variables.

References

Wilson,D. B., &amp Lipsey, M. W. (2001). Practical meta-analysis. Оригиналпрезен тации (см. http://www. mason. gmu. edu/).