Posted: September 1st, 2023

H.W.2. essay

2-15-18 cd/jk/bf
Week 1 Lecture Video for BUS 7106 – Statistics II
In the first week of this course, you will revisit fundamentals of introductory
statistics including topics of hypothesis testing, probability, and a refresher
of how to generate descriptive statistics and correlation analysis using
standard software used in business and graduate school.
Hypothesis testing is a standard research tool used to determine the
statistical significance of research findings. The thinking behind hypothesis
testing is that there is a cutoff, so to speak, of when a statistical finding is
considered to be significant and worthy of further exploration. When a
traditional statistical test (e.g., correlation, t-test, ANOVA, etc.) is
calculated, it reveals a probability value (p-value). This is compared to a
cutoff value to determine this statistical significance. For example, if a cutoff
of 5% is used, then the p-value must be equal to or less than .05 for the
finding to be considered statistically significant (hereafter referred to as
significant). A p-value of greater than this .05 cutoff is considered to be nonsignificant. It is important to note that while a finding is non-significant, it
does not mean that it is not relevant.
The .05 cutoff value, referred to above, is not arbitrary. This cutoff, or alpha,
value is the most commonly used cutoff for significance in research across a
variety of disciplines (e.g., leadership, management, business, the natural
sciences, healthcare, sociology, etc.). The reason this value is common is
that it is the point at which Types I and II error are balanced. If you think
back to your last statistics course, there is always the possibility of
committing an error in hypothesis testing. And to raise or lower the alpha
value will have a resulting impact on Type II value. For example, if you
reduce the alpha cutoff to .01 to demonstrate more stringent evidence of
significance, you increase the risk for the Type II error.
In a review of the types of error, for example, an error may result in a false
positive or false negative. A false positive is thinking that you have a finding,
but in fact it is not real. It is akin to a blood test in which you think you may
have an illness, but the test yielded the false positive. This is known as being
gullible to something that does not exist.
The false negative is not knowing that a phenomenon is in play when it is.
Continuing the previous example, a blood test does not yield evidence that
2-15-18 cd/jk/bf
illness is in effect, but in fact it is. This is being blind. So the question
becomes, would you rather be gullible or blind?
The preceding discussion is relevant to the first Week 1 assignment. While
hypothesis testing has been a standard research determinant of significance
for many decades, it has its criticisms. Two that are related are noted here.
The first is the alpha cutoff. While the .05 value balances the risk of the two
types of error, what if a statistical test yields p =.06? Does this mean that
since it does not meet the alpha cutoff value of .05, and is thus nonsignificant, that it is not somehow meaningful? In effect, the .05 would
indicate that the possibility of a Type I error is 5%. A result of .06 means
that the possibility of a Type I error is 6%. Is the finding, albeit close, not
worthy of discussion? A finding may be non-significant, but still meaningful.
Conversely, given a large enough sample size, a statistically significant
result may not be meaningfully relevant.