When you reply to two your classmates' posts, respectfully agree/disagree to the post and explain why, while offering an opinion with an example to support it and references. Be sure to do one or more of the following:

• Agree and explain why you agree.
• Respectfully disagree and explain why you disagree.
• Offer an opinion with an example to support it.
• Propose an idea or a suggestion.
• Tell a related personal story.
• Respond to the classmates' or the instructor's questions.
• Pose a question and be sure to include your own answer as an example.
• Explain how the discussion question connects to real-life experiences.

Classmate 1:

Hello everyone,

Suppose you are trying to classify a variable where 96% of its observations equal 0 and only 4% equal 1. You run a logistic regression, and the classification table shows that 97% of the classifications are correct. Why might this large percentage still not be cause for celebration?

By predicting 0 for 96% of the observations, the model can achieve high accuracy. Although the overall percentage looks good, the real challenge is to be able to correctly identify the minority class, which only makes up 4% of the data. "The simplest form of logistic regression is binary or binomial logistic regression in which the target or dependent variable can have only 2 possible types either 1 or 0" (Pajput, 2021). It is true that if you run a logistic regression, and the classification table shows that the model is excelling at predicting the majority class but failing to accurately identify the minority class that means that the model that always predicts 0 would achieve 96% accuracy without learning anything useful. The 97% accuracy could just mean the model is taking advantage of this class imbalance rather than truly understanding the patterns in the data. Basically, a high overall accuracy doesn't necessarily mean the model is good at this.

There are other metrics, such as precision and recall for the minority class, that need to be examined. "Precision is a measure of the probability that an event classified by our model with 1 has been correctly classified. So, you want your model to be as correct as possible when it says 1 and don't care too much when it predicts 0" (Malato, 2021). If the model is simply predicting 0 almost every time, it won't be useful for identifying the rare but important instances of 1. "Recall is very used when you have to correctly classify some event that has already occurred" (Malato, 2021). An example of recall would be when trying to detect fraud; we don't necessarily need to know about real 0s, but we want to spot the real 1s as often as possible. So, while 97% accuracy sounds good on the surface, it may not indicate a well-performing or valuable model if the minority class is the real focus.

References:

Malato, G. (2023, October 30). Precision, recall, accuracy. how to choose?. Your Data Teacher. https://www.yourdatateacher.com/2021/06/07/precision-recall-accuracy-how-to-choose/Links to an external site.

Rajput, M. (2021, April 19). Why we need logistic regression?. Medium. https://medium.com/analytics-vidhya/why-we-need-logistic-regression-78f7ee286a3fLinks to an external site.

Classmate 2:

George Lawton tell us that "Logistic regression is a statistical analysis method to predict a binary outcome, such as yes or no, based on prior observations of a data set.(Lawton et al.,2022)While achieving a 97% correct classification rate might initially seem like a reason to celebrate, there are important factors to consider before concluding that the logistic regression model is effective, especially given the highly difference nature of our example dataset.

The scenario shows a big difference from the variables that equal zero (96%) and those that equal one (4%). In situations like this, the classification could end up with a high accuracy by just guessing the most common outcome every time, in our example that would be a variable equal to zero.

The part of the statement where it says that "the classification table shows that 97 of the classification are correct" can be misleading because is telling us that classifying the variable as zero it will 96% correct each time. But this is not necessarily true we are not provided with any supporting information to get the full picture.

Given that we are told that there are two types of classification for our variable, if we only focus on the one that yields the highest percentage, we may lose important information about the rest in our case those classified as one. A good example of this recently happened to me, my son was accepted to our local Chater school recently his teacher plan accommodates the class she has had since the biggening of the school year witch to any other parent this will not present a problem, but to me it a problem because my son's not only just started but also has a history of struggling with science (not his favor subject). Her school year plan caters to 99% of her class but she is neglecting the needs of my son.

In conclusion, and using my son's example while his teacher's class is among the top classes and is recognized by the principal very often, its crucial to consider the needs of all students because unusual cases like the one I am facing with my son can be ignored.

Reference:

Lawton, G., Burns, E., & Rosencrance, L. (2022b, January 20). What is logistic regression? - definition from Searchbusinessanalytics. Business Analytics. https://www.techtarget.com/searchbusinessanalytics/definition/logistic-regression