Data Analysis Flashcards
Where data is complied from
Data used in transport modelling is compiled from samples of the population
Sampling Methods
- Simple Random Sampling
- Stratified Random Sampling
Simple Random Sampling
Involved associating an identifier (number) to each unit in population, then selection numbers at random to obtain the sample
Stratified Random Sampling
Population subdivided into homogeneous strata and then random samples taken from each of these groups
Problem with simple random sampling
Far too large sample would be required to ensure sufficient data collected on minority groups
Types of errors that can be introduced in sampling
- Sampling Error
- Sampling Bias
Sampling Error
Error generated due to fact that sample is only proportion of population
Sampling Bias
Caused by mistakes made either
- when defining population of interest
- when selecting sample method
Equations
In lecture slide 6
Type of errors
- Errors in modelling and forecasting
- measurement errors
- sampling errors
- specification errors
- transfer errors
- aggregation errors
Errors in modelling and forecasting
ideal req is to find combo of model complexity and data accuracy which best fits required forecasting precision + study budget
measurement errors
survey questions badly interpreted, answered badly, coding errors, etc, can cause these
sampling errors
due to representation of population by finite data sets
equation in lecture 6
specification errors
arise where phenomenon being modeled is not well understood, eg. irrelevant variable included in model or relevant variable is omitted
transfer errors
arise if model is removed from one area to another
aggregation errors
typically in models, forecasting done for groups of individuals but data is compiled on basis of responses of individuals
type of info required by surveys
- infrastructure eg. road network, public transport network
- land use inventory eg. residential zones
- O-D travel surveys eg. traffic counts
- Socio-economic info eg. income, car ownership
questionnaire design
- keep qs simple + direct
- divide into several sections
roadside interviews
-better method of estimating trip matrices than home interviews as larger samples available
cordon surveys
provide useful info about external-external and external-internal trips
screen-line surveys
divide area into large natural zones eg. at both sides of river of motorway
travel diary surveys
- require similar but more detail to that of an O-D survey
- diaries distributed to members in a HH and each asked to complete diaries for all travel during day
stated preference surveys
where travelers evaluate and rank set of hypothetical options
longitudinal/time series collection metods
- repeated cross sectional survey
- similar measurements conducted on samples at diff times
- individuals may be included in more than one survey
panel survey
similar measurements made on same sample at diff times
cohort survey
some individuals included for only proportion of survey
problems
- panel surveys become unrepresentative as individuals age
- may omit phenomena eg. children leaving home
- typically higher rate of non-response