Maths and Stats Support

Use: Testing for an association between two categorical variables or for group differences

Dependent (Outcome): Nominal;
Independent (predictor): Two or more groups

Example:Is there an association between newspaper readership and voting preferences.

Summary statistics/graphs:%'s out of the independent variable e.g. what % of Guardian readers voted labour and stacked/multiple bar charts.

Note: If you have ordinal data e.g. strongly disagree- strongly agree responses, the Mann-Whitney test is more appropriate

Use: One nominal variable is tested against population proportion or testing whether the data fits a specific probability distribution.  You will only have one variable and compare to numbered values from the population

Dependent (outcome) variable: Two or more categories 

Example: testing whether eye colour proportions in Sheffield match general UK proportions

Summary statistics/graphs: Percentages and bar/pie chart

Use: Tests which of multiple independent variables are significant predictors of a binary outcome such as survival and produces a model (regression equation) to predict the likelihood of the event happening.

Dependent (Outcome): Binary (2 categories);
Independent (predictor): Any number of continuous or binary variables.

Example: Car insurance companies use logistic regression to identify the factors which increase the likelihood of someone crashing. Car insurance premiums are then based on the predicted probability of you having a crash

Summary statistics/graphs:%'s for binary outcomes and means/ standard deviation for continuous independent variables

Note: If you have more than two categories for your dependent variable, consider combining categories so that there are only two categories and using logistic regression rather than using more complex techniques such as multinomial or ordinal regression