# Unstacked data

Document Sample

```					       Performance Measurement
To get your “money’s worth” from this class,
you should have:

basic calculations
• understood z-scores
• practiced basic formulas and dragging.
• Done some DIY’s
Performance Measurement
Comparing several sets of data

Detecting changes or differences between
2 samples or groups

• Customer satisfaction for males and females
• Weekly complaints last year versus this year
• Number of defectives per shipment using old
process versus new process

Same performance measure – two groups
Performance Measurement
Comparing several sets of data

Detecting changes or differences between
several samples or groups

• Expenditure of different age-groupings
• Default rates and amounts of low, medium and
high risk borrowers.
• Revenue across the different days of the week.
• Profit performance of three or more currency

Same performance measure – several groups
Topics Covered
Excel and data presentation
• Stacking and Unstacking data (AWZ p99, 443)
• Using mean/stdev to compare performance
• Multiple Boxplots            (AWZ 3.8, p443,447)
• z-scores (again)
Some finance concepts
• Risk and return
• Sharp ratio

ILLUSTRATED USING CALL CENTRE DATA AS
WELL AS THE 3 TRADERS CASE.
Stacked and Unstacked Data

Several Samples of data – can come in two
forms: unstacked or stacked

Unstacked data – each data set is in a
separate column (with column labels
describing each sample.
Stacked data – all the data is in a single
column. A second column of labels says
which sample each data point is from.
Stacked and Unstacked Data
Stacked data – all the data is in a
single column. A second column
carries labeling saying which
sample each data point is from.
Take rating as variable and gender
as label.
Unstacked data – each data
set is in a separate column
(with column labels describing
each sample). Here we want
two separate columns of
ratings for M’s and F’s.
Slidesdata/call_centre1.xls

Stacked and Unstacked Data

Unstacked data – each data
set is in a separate column
(with column labels describing
each sample). Here we want
two separate columns of
ratings for M’s and F’s.
Use StatproGo
Data Utilities
Stack Variables…
Unstack Variables…
Topics Covered
Excel and data presentation
• Stacking and Unstacking data
• Using mean/stdev to compare samples
• Multiple Boxplots
• z-scores (again)
Some finance concepts
• Risk and return
• Sharp ratio
Slidesdata/call_centre1B.xls

Comparing Means and Std.Devs
Data distribution is largely summarised by
sample mean and standard deviation
Unstacked data: Use StatproGo then
Summary Stats then One-variable. Select all
variables you wish to summarise. You can also
use ordinary Excel.
Stacked data: Use Pivotables. Break down
data by coding variable. Select Mean and
Std.Dev as summary statistics.
Topics Covered
Excel and data presentation
• Stacking and Unstacking data
• Using mean/stdev to compare samples
• Multiple Boxplots
• z-scores (again)
Some finance concepts
• Risk and return
• Sharp ratio

You have 26 weeks of growth data from three

5 minute breakout
• Use StatproGo to get means and stdev.
• What aspects of performance do the mean and
standard deviation measure?
• Who seems to deserve promotion and why?
• What other statistics might be useful? How
Don’t look at next slides!

You have 26 weeks of growth data from three

Neville       Ann          Brian
Mean       0.242%       0.310%       0.303%
Stdev      0.050%       0.083%       0.017%
Min        0.173%       0.188%       0.262%
Max        0.348%       0.506%       0.320%
Range      0.175%       0.318%       0.058%
Growth     6.490%       8.380%       8.190%

Do we really care about risk (stdev)? Isn’t growth
the bottom line?

Let’s take a vote. Who would you promote?

Open https://qp.e.unimelb.edu.au/clloyd and save to favs

A = Ann
B = Brian
C = Neville
Example. The three currency traders (Class prep chart)

You have 26 weeks of growth data from three

• Are any of these traders any good?
• How could you check?

5 minute breakout
• Who are the best traders? Who are the worst?
• Produce a plot that displays the relative
performance of all the traders. (Remember the
Topics Covered
Excel and data presentation
• Stacking and Unstacking data
• Using mean/stdev to compare samples
• Multiple Boxplots
• z-scores (again)
Some finance concepts
• Risk and return
• Sharp ratio
Big 15 stocks (from class 17)

Sharp ratio
For any investment the return (daily, weekly, monthly)
is uncertain. It has a distribution with two features:
(1) Mean return which you want to be high
(2) Stdev (called volatility) which you want to be low.

• How can you combine risk and return into a single
number?

5 minute breakout
• How good is Brian? He obviously stands out.
How far does he stand out? Give a number!
Topics Covered
Excel and data presentation
• Stacking and Unstacking data
• Using mean/stdev to compare samples
• Multiple Boxplots
• z-scores (again)
Some finance concepts
• Risk and return
• Sharp ratio
Boxplots – A simpler Display

Describes the overall distribution of a set of
numbers but is simpler than a histogram.
Useful when comparing several samples
because too many histograms on one graph
would be both crowded and confusing.
Also produces useful display with small data
sets. Histograms require at least 40 data
values for a reliable picture.
Boxplots – A simpler Display

S=smallest, L=Largest, M=median
Q1=lower quartile, Q3=upper quartile
Watch the Movie!

Boxplots – A simpler Display
Add the sample mean as red diamond.
Points too far from centre displayed
separately (as aberrant points)

You have 26 weeks of growth data from three

Brian’s returns are unusually consistent – the
stdev is very low.
Q: Are there any more subtle differences in the
pattern/shape/profile of Brian, or indeed of the

Neville
Ann
Brian

• Suppose you got a
job betting on the
outcome of the toss
of a fair coin.
• 66 hour week –
1 toss per hour
You Always Bet \$1 on Heads
After a week

that you do not have
any expertise at coin
tossing investment
and suggests that
better employed
elsewhere.
Sitting at the bar..

• Jim, an experienced
play the coin toss
game.
• “Whenever you lose,
double plus one.”
• “When you win, go
back to one.”
How it works:

Bet   Outcome        Payoff                Cumulative
1     Tails          -1                    0
3     Tails          -3                    -3
1     Tails          -1                    3
1     Tails          -1                    7
John Bilson - Financial Risk Management
Weekly record : (11 tosses x 6 days)
Wow!!! Says the Boss
• Out of 66 bets, you
• You sure know how to
pick a coin toss!
• I really liked the way
yourself out of
• Would you like to be
of currency options?
Record over many weeks
Record over many weeks
Record over many weeks
Moral of this Case

If you expect people to make large
consistent sums of money from an
efficient market, you have to
assume that there is a small
probability of a very large loss.
KEY TAKE AWAYS CLASSES 1-3
• Data can come in stacked or unstacked form.
• Comparing data sets
Numeric:
sample mean, std. dev and z-scores can reveal much
Graphical
large data sets:         data dist or histograms
many small data sets: boxplots (any number)
• Excel:
– Stacking and unstacking (StatproGo)
– Pivot tables for comparing samples
– Multiple Boxplots (StatproGo)
– COUNTIF, templates
• Finance
– risk, return, Sharp ratio and RFR

```
DOCUMENT INFO
Shared By:
Categories:
Tags:
Stats:
 views: 0 posted: 7/5/2013 language: Unknown pages: 38
caifeng li