Missing data

Question 1

what is missing data?

Answer 1

Often don’t get information for every measure from every participants
Variables we study may contain missing values/ information
This includes error in data

Question 2

why is missing data important?

Answer 2

Missing data points could be meaningful for analysis if observed
This could potentially affect our analysis as it may conceal a meaningful value for understanding a problem
We need to identify this and determine the nature of the missing data

Question 3

what are the different reasons why missing data may occur?

Answer 3

non-response
data entry errors
instrumentation issues
privacy concerns
natural causes

Question 4

what is non-response?

Answer 4

a reason why missing data may occur if participants choose not to answer certain questions or fail to provide information

Question 5

what is data entry errors?

Answer 5

a reason why missing data may occur if there are mistakes made by researchers during the data collection or entry process

Question 6

what is instrumentation issues?

Answer 6

a reason why missing data may occur if there are problems with the tools or instruments used to collect data

Question 7

what is privacy concerns?

Answer 7

a reason why missing data may occur if sensitive information is omitted from a dataset to protect participants and to maintain ethics

Question 8

what is natural causes?

Answer 8

a reason why missing data may occur if there are events beyond control e.g. technical issues, power outages, environmental factors can lead to missing data

Question 9

what are the different types of missing data?

Answer 9

non-systematic
systematic

Question 10

what are the types of non-systematic missing data?

Answer 10

missing completely at random (MCAR)
missing at random (MAR)

Question 11

what are the different types of systematic missing data?

Answer 11

missing not at random (MNAR)

Question 12

explain what MCAR is

Answer 12

The fact data is missing is independent of the observed and unobserved value
The missing data reduces the analysable population of the study
Reduces the statistical power but does not introduce bias
Can be considered a simple random sample of the full data set

Question 13

explain what MAR is

Answer 13

The missing data is systematically related to the observed but not the unobserved values
Can occur if probability of completion of the survey is related to their sex which is fully observed but not the severity of their depression

Question 14

explain what MNAR is

Answer 14

Missing data is systematically related to the unobserved value
Analysis of a dataset containing MNAR data
Likely to result in biased estimates

Question 15

why is missing data a problem?

Answer 15

Can have significant effects on statistical analyses
May lead to biased estimates
Can affect the story we try to tell about our participants and who they represent in the wider population

Question 16

why do we report missing data?

Answer 16

• Scientific research relies of clear reporting of approach and analysis so we need to show researchers and readers our procedure
• Aid transparency and reproducibility
• Help readers understand potential limitations with our statistical outputs
• Helps to draw reasonable conclusions about our study

Question 17

what missing data do we report?

Answer 17

• The number of participants who had no data on either measure
• The number of participants who had no data on each measure
• Also specify the reasons for the missing data

Question 18

why do we specify the reasons for the missing data?

Answer 18

to provide insights into whether the missingness is random or systematic

Question 19

what are some ways in which we can identify missing data?

Answer 19

• Check within database and scan to look for the missing or erroneous value
• Missing value analysis in SPSS
• Visualise the data and check for the missing data

Question 20

what are some other approaches to missing data?

Answer 20

complete case analysis (CCA)
mean/median/mode imputation
multiple imputation
weighted estimation
model-based methods

Question 21

what is complete case analysis?

Answer 21

• involves excluding cases with missing data from analysis
• Can lead to biased results if missing data is not completely random
• Difficult to make inferences about entire target population or subpopulations

Question 22

what is mean/median/mode imputation?

Answer 22

• Filling missing values with the mean, median or mode of the observed values for that variable
• Generally inappropriate as can reduce the variability of your data
• Can affect measurements of covariances and correlations
• No straightforward way to estimate standard errors

Question 23

what is multiple imputation?

Answer 23

creating, analysing and combing multiple complete datasets with imputed values

Question 24

what is weighted estimation?

Answer 24

using statistical models to estimate missing values

Question 25

what is model-based methods?

Answer 25

using statistical models to estimate missing values

Question 26

why are these other approaches to missing data important?

Answer 26

In addition to describing and reporting missing data, various methods can be used to estimate missing values The choice of method depends on the nature of the data and the study questions