There are different types of data you may want to collect. These are nominal, ordinal, interval, and ratio, and sometimes locational. A nominal data value has no numeric meaning or any particular order, such as green eyes could equal one, blue eyes could equal two, brown eyes could equal three, and so forth. The statistical use would be in accounting or producing a mode, or a consquare distribution. Ordinal data value is where there was no interval between the numbers. But the numbers do represent a rank or an order of sort. For example, types of defects could be A, B, C, and D. Your collection of defect data follows under one of these defect categories. The ranking comes in where a is a major defect all the way to d, where it is a minor defect. For statistical use, the level of defect can be measured and tested using a target value. Also, you can calculate the median, the interquartile range, and it's also used for significance testing. The interval data values are intended to be values, where the scale is equal distance from each other. For example, many of us have filled out questionnaires with highly agree, agree, neutral, disagree, and highly disagree. Since these can be turned into values, such as highly agree equal one, agree equal two, and so forth. Arithmetic calculation can also be perform, such as the mean, the standard deviation, you can have a T test or an F test. The ratio value means there is a zero balance point in the data. An example would be, a number scale with a minus values on the left side, and a zero in the middle, and positive or plus values on the right-hand side. Any distances between these numbers represent a ratio regardless of how high or low the numbers get. So a minus 200 is twice that of a minus 100. And also positive 400 is twice that of a positive 200. For statistics, ratio data can be used in multiplication or division of scale values, geometric mean, or the calculation of the coefficient of variation. There is a fifth type of data called locational data that has a specific purpose, such as its usefulness on a production line to capture the locations where the defects occur. There are two basic type of data, qualitative and quantitative. Qualitative data is not numerical and is limited statistically. Example of qualitative data, is collecting someone's daily activity or asking open ended questions that are answered in an essay form. Face to face interview data, and a sample, like your blood type. Quantitative data are numeric in nature, but is viewed as continuous data or as discrete data. Continuous data are measured on a continuous scale, such as a person's weight or height. It is called continuous, because there is an infinite number of values available. For instance, the weight of a person is typically measured to the nearest whole number, but technically there are infinite number of weights if you use the decimal point. Discrete numbers on the other hand are limited and typical used in counting. For instance, the US Census Bureau counts the number individuals living in a household. On an assembly line, you could count the number of defects. There are a number of ways to collect data. Designing and deploying a survey is a popular choice when large sample sizes are needed. The feedback is quick, and captures first thoughts of the participants. The downside is the response rate is usually low, and the quality of the responses can vary if your questions are not designed well. Face to face interviews and focus groups provide reliable data, because the audience is captive, and you can factor in body language and emotions with your data collection. The downside is that, this method is often more expensive to conduct, and is typically not feasible when you desire large sample sizes. Using e-mail and websites can provide an electronic platform for your participants to either gain more information about a product, or to provide information such including a special sales code on the website. Customer feedback, test marketing, and mystery shopper are methods used to improve the quality of your products. Feedback can be in the form of customer returns, or a 1-800 phone number, to a customer service representative for instance. These complaints can be viewed as constructive criticism. Automatic data collection, are typically monitoring devices designed to focus on a collection of real time data to help with improvements. For example, using high speed barcode readers or vision systems to count, sort, or inspect product as it travels along a conveyor belt. There are other methods of collecting data, these are just a few. Unless you're the US Census Bureau and collect information from everyone, you are likely be using a representative sample to conduct your analysis. There are three types of sampling, random, sequential and stratified. Random sampling is the most common representative of population. The data you collect must have an equal probability of being pick. For example, if I wanted to take a political poll of the entire country, I would not pick one city from my sample set. This would not be representative of the entire country, only for that city. The better way would be to collect samples from as many cities across the country as possible. To avoid bias in picking the cities, it would be wise to use a random number generator, or random number chart to assist you. Sequential sampling is used in destructive and reliability testing, and performed in sequence. The desire is to pass a quality test, and products are used until you achieve the desired result. Stratified sampling is used when products or processes, or mixed, or non- homogeneity exist. This makes it difficult to analyze statistically due to the many factors or variables involved. It is better to break down the processes to single out or stratify the lot based on the machine, or subprocess, or a single lot size. Then pick random samples from each stratified group. Statistical analysis can then be performed.