Data skills - Navigating uncertainty

types of variables

Variables are facets of data that can be split into two categories

• categorical data: qualitative data often recorded as words that sort subjects into groups, rankings or provide labels to describe said subject

• Numeric data: quantitative data recorded as number that can be further analysed through maths

types of catagorical varibles

ordinal: variable with an inherent order to them ( gold, silver, bronze, or agree to disagree)

nominal: variable with no order (colours, species, animals)

types of numerical veraibles

discrete: variables that have set of whole integer values (number of fish caught)

continuous: variables with no set integer values that can keep going forever ( temp, time)

what are subjects in a study

can either be people, objects, observations or cases that are being study and who’s action are measured

what is aggregate data

data that has been gathered from multiple subjects and combined to give averages

What is a lurking variable

A hidden or unintended variable that affects the relationship between the dependent and independent variables

• when this variable is categorical, it can cause Simpson’s paradox

what is simpson’s paradox*

Trends that appear when you look at data in different groups in a data set but disappear or reverse in meaning when the data is looked at as a whole

• Overall, it may seem like more men are accepted into uni than women, but when you look at individual study areas, you see that more women applied to more competitive areas and hence fewer got in.

causation vs correlation

Both describe two different ways a change in one variable can affect another

• causation: when a change in one directly changes another

• correlation: when a variable changes but not necessarily due to a change in another, but a wider variable that affects them both

descriptive vs inferencial stats

Descriptive: summarised/visualised data that display important features of a data set that can be used to draw conclusions

Inferential: using data from smaller groups to conclude larger groups

• not always significant and require a larger sample size to be accurate

what is a designed experiment

When scientists impose a treatment /conditions on subjects and measures their response with the aim of determine a causal relationship

what is selection bias

When the selection of subjects for a study is not representative of a population and instead favours a certain demographic

what is a survey

An observational study that asks participants a series of question of which they self-report

• response rate is determined by the time the survey is sent out and when the survey is completed

• people with stronger opinions are often the ones to reply which can slightly skew the results

what to look at when assesing a study

• study design ( experiment or observational)

• subjects and variables

• Measuring techniques ( method)

• possible sources of bias

what is control in an experiment

In an experimental design, it is important to control as many unwanted variables as possible so that the data gained shows only the relationship between the wanted variables

• often done using a treatment and control groups whose results sre compared to establish causality

• can also be achieved using blinding, randomisation and using blocks

what is the placebo effect

when you tell a group of subjects you are giving them something that may have certain side effect but in fact doesn’t

• called a placebo and must look, smell and taste the same as what you claim it to be

what is blinding

when you don’t tell a participant whether they are in the treatment or control group to prevent bias ( used in the placebo effect)

• double-blinding ( when the subjects and the observers don’t know wh is in which group)

• triple-blinding (subjects, observers and analysers don’t know who is in what group)

why is randomisation of sample important

it allows the unbiased selection of subjects for a experiment that increases the likelihood of a representative sample

• can average out lurking variables

what is a block in an experiment

a group of subjects that have specific characteristics that are thought to effect the dependent variable

• group will all have same characteristic but will be randomly assign to a treatment and control group

what is a matched pair experiment

a form of blocked experiment when two identical experimental units are assigned to either the treatment or control group and their results are compared

• sometimes it is the same subject being placed in both groups as a before and after