need tomorrow
Nokcool
DQStatisticsRepsonsesweek5.docx1. Cameron Izzi
WednesdayOct 14 at 10:25am
When talking about correlation and regression it is important to remember that they are two different statistical techniques used to draw information from data. Correlation is used to measure the relationship of two (or more) variables to some degree. Whereas regression is a technique to see how a variable or variables affects an each other in some fashion. Both techniques have their role and it is important to understand how to use them to gain information on data. If one is trying to draw a conclusion that two variables in a study a closely related then performing a correlation operations on the data to prove that is a great idea. Whereas if one needs to see why a variable is changing or how it is changing they can use a regression technique to find how what is affecting that variable in the the data the most. Each technique is a great statistical methods to exploit data, it is just important to know when to use each one to maximize they strength and value of the method.
WednesdayOct 14 at 6:32pm
Analyzing correlation and regression in research is important because of the identifiable relationships between variables. Furthermore, we are able to analyze the strength and weaknesses of said relationships. We may predict future trends, understand how interrelated the variables are and the effects of changing variable values on the trend or correlation analysis. Correlation analyses are presented on a -1 to 1 scale so we can conclude positive, neutral, and negative correlations of the measured variables. An oft said quote, “One of the first things taught in introductory statistics textbooks is that correlation is not causation. It is also one of the first things forgotten” draws the line on the common conclusion from correlation analyses that lead researchers to align with causal qualities of data (Sowell, 1995). I think it is a great reminder that we must refrain from ascribing qualities of complete certainty to data sets with highly correlated values. Statistical tests allow us the room for inferences instead.
References:
Quotes retrieved from Quotes retrieved from www.goodreads.com/quotes/search?utf8=%E2%9C%93&q=statistics&commit=Search
ThursdayOct 15 at 8:06am
The impact of a study in data form cannot be overlooked. The data tells a concrete story and since most people are visual learners, data represented in graphical form often make the biggest impact when trying to convey what data is trying to say. In looking at the relationship between retail shopping at brick and mortar locations and shopping online, researchers at Entrepreneur began to compile the data and determined that graphical analysis would be the perfect way to convey the impact that online shopping has made, while documenting the shift in consumer habits. Their research looked at different categories like age, gender, countries, and product categories, spending habits, in-store experience, online experience, and what lead to making the purchases. Most of this data was compiled and expressed through scatter plots, bar graphs, pie charts, and frequency polygons. Frequency polygons also show the shape of distribution and are similar to histograms; they consist of line segments connecting the points formed by the intersections of the class midpoint and the class frequencies. This type of plotting helped tell the story of the different categories that the research team used to break down the shopping categories and origin of the shoppers within these two class frequencies. The overall statistical technique that was used in the research was a quantitative analysis of the data that was captured from the categories that were described above. The data was processed based off the variables and represented based on the graphical analysis to be consumed by the audience. The graphical displays from this example demonstrates the overall impact that visual reinforcements can provide from statistical analysis.
References
Leadem, Rose. (2017). 67 Fascinating Facts About E-commerce vs. Brick and Mortar (Infographic). Entrepreneur. Retrieved from https://www.entrepreneur.com/article/306678 (Links to an external site.)
Lind, D. A., Marchal, W. G., & Wathen, S. A. (2017). Statistical techniques in business and economics (17th ed.). Retrieved from http://connect.mheducation.com/class/
ThursdayOct 15 at 6:14pm
I found a research group that used linear regression techniques to predict tensile strength in rubber blends. I found this research interesting because linear regression is often used in statistical practices, whereas measurements of tensile strengths falls under more of an engineering field that usually is done by testing. In the paper the researchers collected data on different tensile strengthens of mixed rubbers, Martinez states, "the input defined a dataset composed of six factors (type of carbon black, defined by their DBPA or IAN values; amount of carbon black; amount of sulphur; amount of accelerants; amount of processing aids; and amount naphtenic oil) and five levels for each factor" (p. 189).
After having the dataset they used 3 different techniques to try to predict the tensile strength of a mixed rubber. These included linear regression, enhanced linear regression, and generalized linear regression (GLM). The differences being enhanced linear regression used gradient boosting techniques and generalized is a more flexible version of a standard linear regression. They used a correlation study to compare the results of the predictions to the actual values. The from the study the results yielded RSME of 25-35% and CORR of 69-72% for validation on the dataset. This is pretty good for a none neural network based method of prediction. It would be interesting to see how a deep learning neural network based method would perform verse this linear regression method.
Reference:
Martinez, R. F., Iturrondobeitia, M., Jimbert, P., & Ibarretxe, J. (2017). Tensile strength prediction of rubber blends using linear regression techniques. 2017 IEEE 4th International Conference on Soft Computing & Machine Intelligence (ISCMI), Soft Computing & Machine Intelligence (ISCMI), 2017 IEEE 4th International Conference On, 188–192. https://doi-org.proxy-library.ashford.edu/10.1109/ISCMI.2017.8279624
