Identify Patterns In Data: Student Rate

Identify Patterns in Data Student’s Name Institutional Affiliation Identify Patterns in Data Identify the dataset and variables and provide context to the research used to collect the data (in an introduction section). On 12th September of 2015, the US Department of Education launched an online tool as an assistant for consumers (most likely students) to compare higher learning institutions based on their costs and value. The administration of the former US President (Barack Obama) spearheaded the launching of the website. The name of the website is College Scorecard. The website consolidates vital information about colleges and universities, such as yearly cost, average student debt, completion or graduation rate, expected job earnings, and employment rate. The College Scorecard is structured to improve transparency by bestowing the power of choice on the public to check out how various schools are serving their students. As diversification of people’s way of life is largely felt in terms of choices and preferences, different families and students may often choose certain schools based on yearly costs, average student rate, employment rate, etc. Over the years, the initial version of the scorecard was discredited since it lacked qualitative information e.g. location of the schools listed. The limitation was misleading and less reliable in guiding a student to their best-fit college. Additionally, criticism arose upon the revelation of inaccurate loan repayment rates for most colleges published on the site (Hanley, 2017, p. 504). The department quickly aims to fix the mistake because it was a ‘coding error’. In 2017, Donald Trump’s administration released an updated version of the site. Today, prospective scholars can create a customized search for their dream college, compare colleges on the framed search and make a decision on which school to choose. The Data.gov website contains datasets accessible and ready for use by the public. In reference to that, the dataset featured in this paper is also accessible and in RDF, CSV, JSON, and XML form on data.gov. The Most Recent Cohort Scorecard Elements dataset has 102 variables both categorical, such as map location and numerical e.g. the OPEID variable. There are several blank entries, NULL and 0 values. Specifically, the selected variable (SAV_ATG) has both numeric and null values. Identify the analysis performed, pros and cons of calculation, and why they are used (in a method section) Data distribution is a fundamental concept in statistics. By definition, a function explaining how all possible data values and their frequency is known as data distribution. Two measures of distribution; one of central tendency (mean and median) and the other measure of spread (standard deviation) describe a distribution. The most common distribution is the standard normal or Z or Gaussian that is defined by a nice bell-shaped curve. Other distributions include Poisson, binomial, Bernoulli, chi-squared, log-norma, etc. Distributions can as well be symmetric like the Z distribution or asymmetric like skewed distributions. Right-skewed distributions have tails to the right with mean greater than the median and often common in biological data. The mean tends to be more susceptible to affect from large values in the data whereas the median is less affected by outliers. Positive skew is best shown by a QQ-plot i.e. quantile-quantile plot with points forming a curve. For the case of normally distributed data, its points form a straight line (Hanley, 2017, p. 505). Left-skewed distributions are the opposite of right-skewed distributions. Most unheard of are underdispersed and overdispersed distributions. An underdispersed distribution is often relative to some expectation with less variable than expected. Such a distribution occurs when ‘extreme’ values are omitted resulting in the removal of the tail and thus reduced variability. A QQ-plot of underdispersed distribution has an S-shaped curve. Overdispersed distributions possess more variable than expected. Example of such distributed data is count data as it has high frequent entries of 0 with average high and few intermediate counts. When plotted on a QQ-plot, the curve is an inverted S. In statistics, there are two types of data i.e. quantitative and qualitative. Divisively, there are categorical and numerical data (Ordonez, 2011, p. 617). Distributions of these two types of data vary. A dependent variable with organized distribution has visible numbers in each group. On the other hand, when that of numerical data is organized, whether in ascending or descending order and put in groups, graphs and charts are used to examine its shape, center, and variability. The distribution of the data vividly describes the variability of observed data. For clarity, variability is accounted for by known/suspected cause or unknown cause alias random noise (Ordonez, 2011, p. 618). Variation in observed data cannot be simply explained and often referred to as randomness. Randomness in data is often attributed to unobserved processes or error in measurement. The result is a pattern-less data. For instance, variation in climate, animal behavior, etc. Randomness tests are tests of randomness that analyze data distribution to check for randomness (Ordonez, 2011, p. 623). In the case of large data, it is essential to select a sample to represent the whole dataset i.e. to make inferences about the whole dataset. Therefore, a random sample is a subset of a population. To avert bias, a sample has to be randomly selected and from a stable population. In R Studio, data is inspected for randomness or patterns using a ‘runs.test’ function under the ‘randsets’ package. Actually, such a test is always on the ‘run’. A run test of randomness often called Geary test is nonparametric (Ribeiro-Oliveira, Santana, Pereira, & Santos, 2018, p. 4). Quick recap; nonparametric tests are tests whose tests of significance do not concern parameters and assumptions are not based on the underlying distribution. The tests assume that the data is not normally distributed (Ribeiro-Oliveira et al., 2018, p. 5). To check if data follows a Gaussian distribution without using graphics methods, one may get the skewness and kurtosis value of the data. Examples of nonparametric tests are such as one-sample sign test, one-sample Wilcoxon signed rank test, Kolmogorov-Smirnov test, and many others. Autocorrelation in data can also be investigated using a runs test. The test assumes: * Constant mean and variance. ...