Analysing Data

What is a scientific hypothesis?

an educated guess about the outcome of a study based on observation and prior information

What is a null hypothesis?

the theory that the variable you are testing has no effect on the outcome of your study

What is an alternative hypothesis?

the theory that there is a directional effect of the variable you are studying on the outcome of your study (opposite of null hypothesis)

What is a hypothesis test?

the process of checking the credibility of a hypothesis using data

What are the possible errors to make with a null hypothesis?

type I and type II

What is a type I error?

rejecting a true null hypothesis

What is a type II error?

accepting a false null hypothesis

What does the p value tell you about hypothesis testing?

the probability that the result of the study is due to chance

What are some ways to represent the results and credibility of a study?

p-value

odds ratio

confidence interval

What is a t-test?

a statistical tool that helps you to decide if two groups are different from each other or if the difference you see is just random chance

What can a statistical model describe the relationship between?

one dependent and one or more independent variables

What is linear regression?

a way of looking at how one variable is related to another using a straight line graph

What assumptions are made when using linear regression?

there is one dependent and one independent variable

the relationship between the 2 variables is well-described by a straight line

the independent variable is measured without error

for a value of independent variable, the dependent data is normally distributed

the dependent data has a similar variance at each level of independent

the residuals/deviations are symmetrically distributed with mean 0 and constant variance

Why would you use ANOVA (analysis of variance) to analyse data?

to test if the predictors in the model significantly explain the variation in the outcome

What 2 things does ANOVA (analysis of variance) tell you?

how much of the variation in the outcome data is explained by the explanatory variable

how much of the variation is left unexplained

What is multiple regression?

a statistical technique that uses 2 or more independent variables to predict or explain the outcome of a single dependent variable

What is the equation for the linear regression model and what do the components represent?

y = a + Bx

y = response variable

x = continuous explanatory variable

a = y intercept

B = line gradient

What is a confounder?

a variable associated with both the outcome and the explanatory/ causative variable

What models can you use to analyse data when you cannot assume a straight line relationship between the two key variables?

generalised linear models

When can you not use linear regression to analyse data?

binary data

proportion data

ordinal data (using numbers but not numerical scores)

multinomial data - more than 2 unordered categories

count data

Why can you not use linear regression to analyse binary data?

model could predict probabilities outside of the range

cannot assume that there is a normal distribution because that’s not possible

What model would you use to analyse binary data?

logistic regression

What is produced by logistic regression?

odds ratios

What test would you perform to test for the association between two categorical variables?

chi-squared test

What are the limitations of the chi-squared test?

it cannot test for an association between more than 2 variables

Why can you not use a linear regression model to analyse count data?

model will predict numbers less than 0 which would not make any sense