Description
Directions: Please provide an initial discussion based on the assignment below, then respond to one peer only per the attached. I will provide additional documents for Module 3, upon Tutor accepting bid. Thanks.Assignment: The purpose of the post problem set discussion boards is to share personal, real-world examples of how the material in the module can be used. Do not come up with contrived examples just to check the box. Instead, spend some time thinking about your life and how some part of the material provided greater insight into issues you are grappling with in work.
Unformatted Attachment Preview
Maggie:
I found this module to be specifically interesting in the context of Human Resources.
Understanding how to create an analysis focused on real-world problems within the
context of my industry has been incredibly valuable. There are a multitude of variables
that would be important to measure, and measuring these variables could potentially
provide insight into the relationship between factors that I previously thought couldn’t
possibly be accurately measured.
This section also demonstrated how I can better interpret analysis findings for my
organization. In part 6 of this module, we see that “just because you find a result in your
sample doesn’t mean that the true effect is that. In fact, you may find this estimated
effect even when in truth there really is NO relationship whatsoever between the two”.
This point stood out to me because I don’t feel that I have been able to fully understand
the difference between correlation and causation within my specific role up until this
point. I have been making determinations based on data that was presented without
knowing how to effectively translate the variables within the data. My hope is that I can
take the foundational knowledge within this module and use it more effectively within
my role in the future.
Badrina:
The use of statistical regression methods to estimate quantitative data like salaries is a
very relevant use case for the domain I am associated with. While there are no quantities
of parts or widgets that we are estimating in the domain I work in, but figures like salaries
and different factors that affect salaries for different roles and using real-world data to
estimate them is a very useful application of the concepts learned from this
micro/macroeconomics course, and can be used at my firm/company. We need accurate
salary data to submit competitive bids, and this is a very useful application of the
concepts learned here. Every industry will have to deal with it’s own nuances of course,
on what drives demand for their product/service, and is affected both by the inputs (raw
materials and resources) and outputs (services, physical components produced, etc) , and
is measured based both on the costs, and the revenues/profitability for that specific
domain/industry, and the specific firm. The competitive landscape and the capabilities of
the firm (strengths, weaknesses, etc), will ulitmately determine how successful a firm will
be with
Julian:
I am in the midst of developing a public awareness marketing campaign with several
elements that have costs associated with them including creating a microsite/landing
page, paid search engine optimization advertising, digital (display and social) advertising,
broadcast advertising, and earned media, among other tactics. In building this plan, I find
myself evaluating the price tags associated with each item as well as the project costs per
action of each line item expense. What is the cost per click and cost per impression on
display and social advertising? What is the cost per view on broadcast? And so on. Also,
even if penetration and saturation targets are achieved, what is the magnitude of the
causal relationship on those investments and our desired outcome, which is not directly
financial (we are not a sales organization, we are a service organization)? Our desired
outcome is focused on issue persuasion. We will be able to tell whether the outcome we
hope for is realized. The question then is determining through a method like regression
analysis the degree to which our investments may have a causal relationship on that
outcome.
Juan:
This module provided very interesting applications to both my professional and personal
life. In healthcare, we are always trying to figure out what factors have the largest impact
on patient outcomes. When discussing physical therapy specifically, we can use a
regression analysis to understand what factors drive patient improvement when holding
everything else constant. Additionally, I help manage a couple of physical therapy clinics
and would love to know what truly drives compensation and salaries for physical and
occupational therapists. You would think specialties and type of degree (Doctorate vs
Masters) lead to higher salaries, but from my working experience, it seems that years of
experience and setting are the only things considered.
I also had some personal experience utilizing regression analysis when deciding to
purchase a new car. We looked at how different independent variables such as year,
make, model, size, and horsepower can influence the price. I hope to use the information
learned in this module to make better informed decisions in both my personal and
professional life.
Sarah:
I actually just recently used regression analysis to study the impact of many different
variables on post operative patient’s nausea and vomiting. There is a new screen tool we
implemented as a system which comes with specific interventions (mostly medications)
dependent on the risk score. We looked at many variables such as type of surgery,
gender, age, preoperative medications, intraoperative medications, and length of postop
stay to see what impact using this new tool/risk score had on our patients. It was
interesting to note that the score did not have strong relationships to
actual nausea/vomiting post operatively, but there were strong relationships with age
and gender. There was also strong correlation with specific nausea medications and
increased recovery time, which were used less after this intervention was implemented
on the units. Overall it was a small sample size so the ability to generalize the data is low,
but it was very interesting to see how our population was impacted by this simple
screening tool.
Module 3: Part 1
To begin our study of empirical analysis, this part lays out our plan of attack. Try to
keep in mind this broad perspective as we dig into the weeds of statistical tools.
This part covers the following Module learning objectives:
•
State and explain the value of data driven decision-making
Here is the notes file:
There are five general parts to empirical work:
1. Data collection/assessment
2. Describing the simple patterns in the data
a. For variables individually
b. For sets of variables
3. Estimating relationships between variables
4. Evaluating the quality of the empirical work
5. Using the empirical work for decision-making/forecasting
The best way to learn econometrics is through the use of a running example.
I will build the tools of statistical analysis – particularly the third part – with a hands on
example. Then we will apply this to demand data and match up this module with the first
two.
Example: Determining the value of a home:
Suppose that I just sold two homes A, and B, and I want you to guess which home had
the higher price.
You can’t ask me what the prices were, and you can’t ask me what nearby homes
recently sold for.
What information do you want to know?
Characteristics
Yearbuilt
Squarefoot
Lotsize
Baths
House A
85
2700
¼ acre
2.5
House B
90
3100
2 acres
3.5
Distance
Selling price
2 miles
$150,000
7 miles
$210,000
Now, if we could do this for the entire “population” i.e., have data on every house that
was ever sold, we could get a really really accurate understanding of the impact of
different characteristics on the selling price of a home.
Problem, we only can get a sample of data – a small slice of the entire population, from
which we can try to obtain an understanding of the relationship between the independent
variables and the dependent variable.
Plan of attack:
1) look at one variable assuming we actually do have population data.
2) See how our work changes when we only have a sample (how far from the ideal
to we have to go?)
3) Look at all variables simultaneously
4) Perform hypothesis testing.
In practice, we know that when you want to sell your house the realtor will bring up
comparable homes that have recently sold and based on this handful of observations
ballpark an asking price for your home. But can this be done more accurately – and with
more data?
Module 3: Part 2
This part introduces the set-up of the problem we will be working through – real data in
all – to build our econometrics/data analysis toolkit.
This part covers the following module level learning objectives:
•
Define and graph a Population Regression Function
Here is the notes file:
We will start by addressing a simple question: What determines the sale price of a home?
Of course the answer is – just like a demand function – “lots of things”
But the purpose of empirical work is two-fold:
1. Sometimes what you think matters doesn’t, and what you don’t think matters
does.
2. In practice we often care about magnitudes – what matters most and to what
degree?
Some definitions to start:
Dependent Variable: This is the variable we want to understand/explain
In this case, it the sale price of a home: P
So our goal is to understand variations in the price of a home
Independent Variables: These are the variables that explain variations in the dependent
variable.
In this case we have lots of potential variables: size of home, location, age, location,
number of bathrooms, location…
To begin, we will build our econometric tools by focusing on the relationship between
the dependent variable and just 1 independent variable.
Let’s examine the relationship between price and size of house measured in square
footage.
So if you had data on prices of homes and their size what would you first do with the
data?
My hope is you would first just eyeball the data and see if your hunch is correct – larger
homes have larger prices. But you will also find that this isn’t always true: there will be
some (a few) smaller homes that sell for more than a larger home. But, I suspect, you
will see that these are exceptions to the general rule of a positive relationship: the bigger
the home the bigger the price.
Keep this is mind: we are interested in general relationships. There will ALWAYS be
exceptions – do not confuse exceptions with evidence that there isn’t a relationship.
For example, there will exist some high school drop-outs who make more money than
college graduates, but that is the exception. In general, more education tends to lead to
higher wages… You will see people who “cherry pick” the data make this mistake.
Now, after a visual inspection of the data, you would likely plot the data: perform a
simple data visualization using a scatterplot.
With this, we can begin to think more carefully about statistical ideas. To do so, we need
some formal definitions:
Definitions: E(P|sqft = 2000) = conditional mean price of a home that has 2000 square
footage.
Population regression function: PRF
Idea: this is a function that represents the true relationship between the dependent
variable (price) and independent variable (square footage).
Graphically, we have:
Thus, the PRF is the set of all conditional means of the dependent variable over all values
of the independent variable. For this reason, the PRF is often called the conditional mean
function.
Mathematically, we have:
PRF: E(P|sqft) = 0 + 1 sqft
Where: 0 = intercept of the PRF
1 = slope of the PRF
Our goal is to find 0 and 1
Especially 1, this is the TRUE effect of a 1 unit change in square footage on the
AVERAGE price of a home.
So think about this – our work isn’t trying to determine what the relationship is for
your specific home. We know we cannot do this – would love to but can’t….
Instead, we can find the AVERAGE effect of square foot on price and this would be
our best guess for your particular home.
Two important assumptions so far:
1) the effect of increasing square footage doesn’t depend on whether the house is big
or small. That is, we are assuming a linear relationship between the independent
and dependent variable.
2) I drew the distribution of prices for different square foot levels the same. This
may not be the case.
Module 3: Part 3
Part 3 begins to transition from the population analysis to sample analysis. As such we
are moving towards estimation: The population shows the TRUE effect of X on average
Y. Work in the sample will give us an ESTIMATE of this true effect.
This part covers the following module level learning objectives:
•
•
Define and interpret the population regression function intercept and slope
define an error term
The notes are here:
Recall, we are thinking about the relationship between sale price and square foot of a
house.
We are imagining that we have data on every house ever – we have population data –
which will allow us to determine the TRUE relationship between square footage and the
AVERAGE price of a home.
Now, let’s look at a particular house within this framework:
Pi = E(P|sqft = sqfti) + i
Or,
Pi = 0 + 1sqfti + I
So the first part: 0 + 1sqfti is the average price of a home of this particular square
footage.
I is how this particular home’s (i) price is different from the average.
Determinants of I:
1). Most importantly other factors that influence the price of a home!
2). Pure randomness/luck
idea of I
think of this random variable as a trash can that contains all other variables that affect the
dependent variable plus all the random noise that can, in a wholly
unpredictable/unsystematic way influence price.
Much of causal empirical work focuses on the error term. We want the end result
to have an error term that contains purely random noise – nothing of importance in
it.
Summary:
Our goal is to find the population parameters 0 and 1
Since we rarely have population data, we need to work with the next best thing, a sample
of data.
Next step: draw a sample and find a formula that will draw a line that will be a close
approximation of the PRF.
To emphasize: there is a theoretical idea: there is some true average effect of a
change in square footage on price. That is beta 1. The problem is that we will never
know this, so instead we will collect a sample of data from which we can get an
estimate of this true effect (and hope this estimate closely resembles the true effect)
Purchase answer to see full
attachment
