6.4. Data preparation

The raw data collected in the survey need to be processed before direct application in the model. This is necessary, because of various errors, except in the survey both in the selection of sample houses as well as error in filling details. In this section, we will discuss three aspects of data preparation; data correction, data expansion, and data validation.

Before using the survey data in transportation models, it’s crucial to prepare the data to ensure its accuracy and relevance. This preparation helps identify and correct errors that may have occurred during data collection. The main areas of focus in this process include correcting any inaccuracies (data correction), expanding the data to reflect the entire population (data expansion), and validating the collected data to ensure it is consistent and logical (data validation).

Think of data preparation like preparing a recipe before cooking. You can't just throw in random ingredients; you need to measure them accurately (data correction), ensure you have enough ingredients for the number of people you are serving (data expansion), and check that all ingredients are fresh and safe to use (data validation).

Various studies have identified few important errors that need to be corrected, and are listed below.

1. Household size correctionIt may be possible that while choosing the random samples, one may choose either larger or smaller than the average size of the population as observed in the census data and correction should be made accordingly.

2. Socio-demographic corrections It is possible that there may be differences between the distribution of the variables sex, age, etc. between the survey, and the population as observed from the census data. This correction is done after the household size correction.

3. Non-response correction It is possible that there may not be a response from many respondents, possible because they are on travel every day. Corrections should be made to accommodate this, after the previous two corrections.

4. Non-reported trip correction In many survey people underestimate the non-mandatory trips and the actual trips will be much higher than the reported ones. Appropriate correction need to be applied for this.

Data correction is a crucial step that involves identifying and correcting specific types of errors that might affect the accuracy of the data:

1. Household size correction ensures that the sample reflects the average household size from census data.
2. Socio-demographic correction adjusts for potential discrepancies in sex and age distributions.
3. Non-response correction addresses the issue of individuals who didn't respond, ensuring their absence doesn't overly skew the data.
4. Non-reported trip correction accounts for trips that individuals forget to mention, acknowledging that actual travel might be higher than reported.

Imagine you are conducting a survey about the favorite ice cream flavors of students in a large school. If you mistakenly surveyed mainly families of large students, the data would show a preference for ‘double chocolate chip’ when in reality, it's because you surveyed those who love large sizes. Correcting the household size ensures that you get a balanced view that reflects every kind of student. Similarly, if some students don’t respond because they are on vacation, that’s the non-response issue we need to account for in our analysis.

The second step in the data preparation is to amplify the survey data in order to represent the total population of the zone. This is done with the help of an expansion factor which is defined as the ratio of the total number of households addressed in the population to that of the surveyed. A simple expansion factor F for the zone i could be of the following form. a F = (6.1) b d − where a is the total number of households in the original population list, b is the total number of addresses selected as the original sample, and d is the number of samples where no response was obtained.

Sample expansion is the process of adjusting the survey results so that they can be interpreted as representing the entire population. Using an expansion factor helps in scaling the data. For example, if there are 100 total households, but only 20 responded, and 5 didn’t respond, the expansion factor helps calculate how to adjust the findings from those 20 responses to represent all 100 households. This ensures that the analysis accurately reflects the wider population from which the sample was drawn.

Think of sample expansion like blowing up a balloon. You start with a small balloon (your survey sample) that doesn't represent the full size of the balloon (the whole population). By using a pump (the expansion factor), you can make the small balloon bigger until it fully represents the actual size of what you want to study. This way, the insights you gain from the small sample reflect the overall preferences of the entire group.

In order to have confidence on the data collected from a sample population, three validation tests are adopted usually. The first simply considers the consistency of the data by a field visit normally done after data entry stage. The second validation is done by choosing a computational check of the variables. For example, if age of a person is shown some high unrealistic values like 150 years. The last is a logical check done for the internal consistency of the data. For example, if the age of a person is less than 18 years, then he cannot have a driving license. Once these corrections are done, the data is ready to be used in modeling.

Validation of results is crucial to ensure that the processed data is reliable and usable. It involves three key checks:

1. A field visit confirms data consistency and accuracy post-data entry.
2. A computational check scrutinizes the data for unrealistic values, such as an age of 150, which needs to be corrected.
3. Logical checks verify that the data aligns with expected norms, such as confirming that no one younger than 18 would have a driving license. These steps ensure that when the data is used in modeling, it is accurate.

Consider validation like proofreading a paper before submitting it. You carefully check for grammar (data consistency), ensure the numbers add correctly (computational check), and confirm that the content flows logically (logical check). If you skip this step, you might accidentally submit a paper full of errors, just like using unvalidated data can lead to incorrect conclusions in your study.

--