Monday, January 23, 2017

Fall 2016 License Plate Game Report

Fall 2016 License Plate Game Report
by Mark Kloha
© 2016

The Hypothesis
The probability, frequency, and difficulty of finding a license plate from any given state is related to three variables – population, distance, and per capita income for that state.  The number of license plates that I find from any given state will be highly correlated to a combination of these three variables.
Background
I did this type of study in Summer 2016 and Summer 2015.  This is the first time I did this type of study during the Fall.
Methodology
Timing
I began on Wednesday, September 7, and went through Tuesday, January 3 (After Labor Day weekend to after New Year).  For the purposes of this research, each ‘week’ began on Wednesday and ended on the following Tuesday.  This timing allows the weekends to be grouped together.  This study was for 17 weeks.  The Summer 2016 report was for 15 weeks. 

Tools –
When tracking license plates, I will be using an iphone app to track where and when I see a license plate.  The app I am using is “License Plate Zone”.  This app allows me to log any state and any license plate multiple times.  Most license plate apps will only let me log a state once.  This app lets me log multiple license plates for each state. 
Trish does most of the driving on the weekends and in the evenings.  I will be able to log the license plates quite effectively.  If I am driving and see an out-of-state license plate, I can press a button on my iphone to talk to Siri through a Bluetooth-hands-free connection and have Siri make a note of what license plate I just saw.  Then I can enter the information into the app later on.
How I’m Counting
I am only counting the official 50 States in the United States of America. 
I do not track the District of Columbia or any other US territories.
I am only counting the license plate if it is on the back of the car.  Some states require a license plate on the front and back.  Some states only require a license plate on the back of the car.  It is possible for a car to have two different plates from two different states.  This can happen if a person lived in say Hawaii for a while, brought their license plate back (or even their car) to a state that only requires a license plate on the back, and then they kept the license plate on the front.  I have seen a number of Hawaii license plates on the front of a car with a different plate on the back – while these front Hawaii plates are rare in of themselves, these front license plates will not be counted.  I cannot make an exception for Hawaii because then I would need to make an exception for all front license plates.  Technically, counting the front license plate would increase the population of that state to anyone who had ever lived there and just happened to keep their license plate as a memento.  I have no way of adjusting the population factor to accommodate these front license plates.
License plates from semi-trucks, U-Hauls, etc. do not count.  Depending on the state laws, it is more beneficial for certain types of truck companies to be registered in various states.  I’ve seen a lot of semi-trucks with Maine license plates but very few passenger vehicles with a Maine license plate.
I will do my best not to double count license plates.  For example, on my way into work, I see a car parked on the side of the road and it has a Tennessee license plate.  If I see this car on my way into work every day, I will not count it again and again and again.  If I’m at a campground, it is possible that there are campers from out of state there.  As we move around the campsite, I will not record a license plate every time I see the same vehicle again later on. 
Where I’m Counting –
I am only counting out of state license plates that I find in Michigan.  Our summer travel plans are mostly in Michigan.  We have several weekend camping trips planned throughout Michigan.  Also, I will be looking for license plates just in our daily routines.  We will be going to Columbus, Ohio for one weekend.  While out of Michigan, the license plates that I find while out of state will not count. 
Variables -
I have three variables – population, distance, and income with cost-of-living-adjustment (COLA).
Population
The population data is from:
This is for July 2015.
Per Capita Median Income and Cost of Living
The economic data that I use for the Per Capita Median Income variable is from:
This data source provides the per capita median income for each state, the average state taxes on that income, and then the Cost of Living Adjustment factor. 
For the statistical analysis, I took the median income, subtracted the state taxes, and adjusted that based on the COLA percentages for each state. 
(Median Income – State Taxes)/(Cost of Living Factor)

Distance
The third variable is distance.  The app I am using lists the latitude and longitude of where I found that license plate.  For each observation, I calculated the distance from where I saw that license plate to the state’s largest metropolitan area.  Then for each state, I calculated an average distance to that state’s largest metro area.  In my Summer 2015 License Plate report, I tracked distances from both the largest city and also to the state’s border.  The conclusion from that report is that statistically it does not make a difference.
Shortcuts through Canada
To calculate the distance to the New England states, it is quicker and shorter to drive through Canada, and so my calculations for distance did utilize this shortcut.  This is different from last year’s report where I did not allow the shortcut through Canada to be used to get from Michigan to the New England states.
Also, to drive from Alaska to Michigan, it is necessary to drive through Canada as well.
Michigan Ferries across Lake Michigan
To get from Michigan to Wisconsin, Minnesota, or other western states, there are two car ferries that go across Lake Michigan.  The Lake Express goes from Milwaukee to Muskegon in 2.5 hours.  The S.S Badger goes from Manitowoc, WI to Ludington, MI in 4 hours.  When calculating routes in Google Maps, Google Maps always said the quickest way to some places was via ferry - specifically the Lake Express
To calculate the distances to Wisconsin and Minnesota, the distance calculations that I used were based on driving around Lake Michigan, through Chicago.  I did not utilize either of the ferries that go across Lake Michigan.
Distance to Hawaii Calculation
It is possible to find a car with a Hawaii license plate on the back in the mainland and even right here in Michigan.  I did see one this summer!!  It is obviously impossible to drive to Hawaii.  Hawaii has a population and per capita median income but no drivable distance to Michigan.  If I were to include Hawaii in this study, then how should the distance be calculated?  If I use the actual distance from Hawaii to Michigan of 4,500 miles, then this assumes that the distance is drivable – which it isn’t.
It is possible to transport a car from Hawaii to California by boat.  It costs approximately a thousand dollars (give or take a few hundred dollars) and takes ten days.  http://www.matson.com/pov/booking/shipping_rates.htm
There are a few possibilities for dealing with Hawaii:
1.        Not include Hawaii in the study
2.       Convert all the distances to a “time” variable. 
3.       Convert the shipping time and costs from Hawaii to California to a “distance”. 
I originally was not going to include Hawaii in the study; however, I actually did see a license plate from Hawaii on the back of a Jeep, and so I went about converting the traveling time to a calculated distance.
I came up with a method to convert the time at sea to a driving time.  It takes 10 days to ship the car.  The trip will take 10 days, and assuming that an average driver could easily drive 500 miles in one day, then that means the entire trip has been assigned a mileage of 5,000 miles from Hawaii to California, and this will get the vehicle from Honolulu, Hawaii to Los Angeles, California.  The distance from Detroit, MI, to Los Angeles, CA, is 2,218 miles.  It also costs $1,000.  So, adding the converted time on the boat to miles with the actual mile from Michigan to California and adding the cost as miles, this gives 8,218 miles. 
However, with such a large distance of 8,218 miles, this distance makes Hawaii an outlier in the multiple regressions, and so for the in-depth statistical analysis, Hawaii will not be included in the Correlations and Multiple Regressions later on.
Quick Summary
1.       I am only tracking the official 50 States – not including D.C or other U.S. territories.
2.       I am not tracking Michigan.
3.       Distances are measured by the shortest distance from the point I saw the license plate to the largest metro in the other state.
4.       Distances do use the short cut route through Canada to get to the New England States.
5.       The distance to Alaska is based on driving through Canada.
6.       Only license plates on the back of a vehicle count
7.       License plates on semi-trucks and rentable trailers/trucks do not count.
8.       I will be tracking the frequency, date, time, and location for the out of state license plates that I find.

Results:
During the 17 weeks, I traveled, 7,854 miles miles within Michigan (Summer 2016 was 7,023 miles).  I logged 794 out of state license plates (Summer 2016 was 1,031 license plates).  So, the Fall 2016 study was for 2 weeks longer, I traveled 800+ more miles, and I saw 237 fewer license plates.
The top 5 were:
Fall 2016                               Summer 2016
Illinois – 118                        Illinois -195
Ohio – 96                             Ohio - 123
Indiana – 67                        Indiana - 115
Florida – 66                         Florida - 88
Texas – 42                           Texas - 38

These 5 states made up just a little over 50% of my sightings. 

The states that I saw the fewest of:
Utah – 1
New Hampshire - 1
Delaware – 1
Rhode Island – 1
North Dakota – 1
Hawaii – 1
West Virginia - 0


As compared to -
Summer 2016 bottom 5:
Maine -1
Rhode Island -1
South Dakota - 1
Hawaii - 1
Delaware - 0

Other interesting information:


License Plate count by week –

Week Number
Mileage
Observations that week
Week Start Date
Week 1
531
39
Wednesday, September 07, 2016
Week 2
343
44
Wednesday, September 14, 2016
Week 3
425
34
Wednesday, September 21, 2016
Week 4
509
63
Wednesday, September 28, 2016
Week 5
508
55
Wednesday, October 05, 2016
Week 6
478
45
Wednesday, October 12, 2016
Week 7
523
46
Wednesday, October 19, 2016
Week 8
508
31
Wednesday, October 26, 2016
Week 9
514
59
Wednesday, November 02, 2016
Week 10
624
90
Wednesday, November 09, 2016
Week 11
385
42
Wednesday, November 16, 2016
Week 12
443
84
Wednesday, November 23, 2016
Week 13
441
36
Wednesday, November 30, 2016
Week 14
355
26
Wednesday, December 07, 2016
Week 15
398
9
Wednesday, December 14, 2016
Week 16
467
39
Wednesday, December 21, 2016
Week 17
402
52
Wednesday, December 28, 2016




Correlations:
For the following statistical computations, the independent variable is Frequency - how many times I saw each state.  The three dependent variables as mentioned earlier are population, distance, and per capita income adjusted. 
Using Microsoft Excel, I ran independent correlations between the following four variables and the dependent variable.  Here are the results –
Population:  50% correlation
Distance:  30% inverse correlation
Per Capita Income (PCI):  0% correlation
Per Capita Income Adjusted (PCIA):  26% correlation

So, population is the most significant variable.  The distance factor is an inverse correlation.  The inverse correlation means that the closer the state is, then the higher the frequency.  I found it interesting that I got a 0% correlation on the median income but then it jumped to 27% once I adjusted the state incomes based on the Cost of Living Adjustment factor.
Multiple Regressions
A short general summary of multiple regressions.
The Multiple R value is how well did my variables correlate to the number of observations for each state.
The R Squared is the value of the Multiple R value and squared.  The Adjusted R Squared value modifies the R Squared value based on the number of independent variables that are being used in the regression.  The R Squared and the Adjusted R Squared measure how well the three variable explain the changes in the Frequency variable.
When dealing with human behavior, and tracking license plates is human behavior, it is rather typical for the Multiple R value and the R Squared values to be low.  If the R Squared value is above .3, then you’re onto something.  If the R Squared value is above .5, then that is a smashing success.
Somewhat more important are the t-stat values.  This indicates the strength of the variables in the regression.  The T-stat value should be evaluated independently of the R Squared values.  It is possible to have low R Squared values but if the t-stat values are greater than 2.00, then the variables are considered significant, and the variables still affect the dependent variable. 
Even quicker summary – R Squared values above .3 are good, R Squared values above .5 are excellent, and t-stat values above 2.00 are awesome regardless of the R Squared value.





Using Microsoft Excel, I ran several multiple regressions.
(PCIA – Per Capita Income Adjusted)

Population, Distance, and PCIA without Hawaii  - 48 states
Multiple R:  .646
R Squared:  .418
Adjusted R Squared: .378

Pop t-stat:  4.55
Distance t-stat: -2.67
PCIA t-stat:  2.09

This looks at 48 states for the 3 variables – Population, Distance, and PCIA.  The Adjusted R Squared value is above .3, and all three variables are significant. 
Population and Distance without Hawaii – 48 states
Multiple R:  .600
R Squared:  .360
Adjusted R Squared:  .331

Population t-stat:  4.366
Distance t-stat:  -2.78

This looks at 48 states with only two variables – Population and Distance. 
The Adjusted R-squared value is above .3, and the two variables are significant.

Residuals

For, the next set of regressions that I ran, I removed what are called the “Residuals”.   The Residuals in a multiple regression are essentially throwing the data set “off balance”.  If one removes these residuals and analyzes those separately, one can get a better idea of how well the variables can in fact predict the outcome – how well do Population, Distance, and Income predict how many observations I will see from each state without these anomalies/residuals.

The Microsoft Excel Regression analysis showed that Illinois, Indiana, Ohio, Florida, Alaska, and Oregon were residuals.  So, for now, I will detail the results of the multiple regression without these states and re-visit these residual states for a closer look.

Population, Distance, and PCIA (42 states)
Multiple R:  .846
R Squared:  .716
Adjusted R Squared:  .693

Population t-stat:  9.501
Distance t-stat:     -4.237
PCIA – 2.194

Here we see a huge jump in the Multiple R values and the R Squared values.  There is an Adjusted R Squared value of .693.  That means that these variables can decently predict the frequency of seeing out of state license plates rather well.  The T-stat values are all above 2.0.  The fact that Population t-stat is 9.5 indicates that this is the most important factor, with Distance next, and the PCIA income variable contributing to the prediction.

Population and Distance with 42 states:
Multiple R:          .824
R Square:             .680
Adjusted R:        .663

Population t-stat:  8.93
Distance t-stat:  -3.88

Again, the Multiple R value is fairly high, the Adjusted R value is above .5 which is excellent, and the t-stat values are both above 2.0.  The Population T-stat variable is higher which means it’s a little more important than the Distance variable.

Residuals Review:
Let’s get back to those residuals – Illinois, Indiana, Ohio, Florida, Alaska, and Oregon. 

The first five were categorized as residuals as I had seen so many of these that the frequency of these states compared with all the other states was off-balancing the regression model.  Illinois, Indiana, and Ohio – these are the three closest states to Michigan’s southern border.  Florida almost acts like a border state with 66 sightings – Florida was almost tied with Indiana at 67! 

In my Summer 2016 report, I thought that the large sightings of Florida license plates was due to the fact that they were snowbirds.  They might be snowbirds but they haven’t all left.  The number of Florida plates seen could be just a factor of their population.  As with these studies the most frequently seen license plates are either very close to Michigan or one of the more populated states – California, New York, Texas, and Florida.

Alaska – I saw 3 Alaskan license plates.   Last summer I saw 9.  Three license plates from Alaska though was enough to through the regression off-balance still and have it registered as a residual.

Oregon – this showed up as a residual - again.  So, let’s take a look at the data.  I saw 6 license plates from Oregon.  Doesn’t seem like a big deal.  Oregon has a population of about 4,000,000 people - which is about average.  States with a population around or below 1,000,000 people – those are difficult to find.  Oregon has 4,000,000 people so what’s going on? 

The distance factor!!  And their economy!!  Distance factor first.  Portland, Oregon is farther away than Seattle, Washington.  I measured my distance variable from where I saw the license plate to the most populated metro area in the other state.  For Oregon, that is Portland.  To drive to Oregon, one needs to navigate around the mountains in Idaho.  You can either go north of them and drive through Seattle, Washington first and then go to Portland, or you can drive way south of the mountains in Idaho which means driving to Salt Lake City, Utah first then to Boise, Idaho then to Portland, Oregon.  So, of the 48 contiguous states – Oregon is the farthest away.  So, seeing 7 license plates from there is really high.

Income – Oregon has a median income of about $37,000 – this is a little above average.  However, they have a 129% Cost-of-living-adjustment.  So, once when state taxes are taken out and then that amount is adjusted, Oregon has an adjusted median income that comes in 49th place (Hawaii is 50th).

More Regressions

So far, the regressions have been based on a sample size of either 48 states or 42 states.  To increase the sample size, I employed a new method.  Each state for each week is its own sample.  For example, there is Alabama1, Alabama2, Alabama3……Alabama17; Alaska1, Alaska2, Alaska3……Alaska17.  So, there are 48 states over 17 weeks which yields a sample size of 816.

The first regression for all 816 States per Week:
Multiple R                           0.446
R Square                              0.199
Adjusted R Square          0.196

Population t-stat              11.479
Distance t-stat                   -6.784
PCIA t-stat                          5.462

Here the R Square and the Adjusted R Square values are very low.  The three variables however are showing as being significant.  Remember that the t-stat values need to be evaluated separately from the R Square values.

Second regression for 814 States per Week:
Multiple R                           0.521
R Square                              0.271
Adjusted R Square          0.269

Population                          14.994
Distance t-stat                  -7.374
PCIA t-stat                          5.759

In this regression, I took out only 4 residuals – Illinois10, Ohio10, Illinois12, and Indiana 12.  The number of sightings for these states was higher than normal. 

With this regression, the Multiple R, R Square, and Adjusted R Square values all increase considerably.  The three variables with a t-stat value well above 2.0, this shows again that these three variables are significant in determining the number of out of state license plates one can reasonably expect to see here in Michigan based on the three variables.


New Variables

In the world of academic Geography, they like to be all scientific and stuff and they have various equations to calculate interactions between two places.  It’s based on Newton’s law of gravitational attraction F=m1*m2/d2.  Gravitational Force is equal to the mass of object 1 times mass of object 2 divided by the distance squared.  This equation has been adopted by geographers to the following:

Demographic Gravitation is:

DG = p1*p2/d2

Demographic gravitation is equal to the population of place 1 times the population of place 2 divided by the distance squared. 

The other equation is Demographic Energy:

DE=p1*p2/d

Demographic Energy is equal to the population of place 1 times the population of place 2 divided by the distance.

For the purposes of this research, I have adapted these equations further.  Since I was focusing just on the license plates found here in Michigan, it is not necessary to multiply the population of the other states by Michigan’s population 49 times.

The three new variables to measure Demographic Interaction between the other states and Michigan are:

                P/D = Population/Distance
                (P*IA)/D = (Population*Income Adjusted)/Distance
                PIA= Population*Income Adjusted

I ran the correlation between my new variables and frequency for 48 states (without Hawaii or Michigan):

                Population/Distance =                                                94% correlation
                Population*Income Adjusted/Distance =                  93% correlation
                Population*Income Adjusted =                                 55% correlation

For the Population, Distance, Income composite variable (PDI), I ran some more multiple regressions.

PDI
Multiple R                            0.940
R Square                              0.894
Adjusted R Square               0.881

PDI t-stat                             18.76

So, this is showing the Multiple R value at .94 with the R Square and Adjusted R Square values very close together at .8941 and .881 respectively. 

Furthermore, in this regression, the residual analysis showed that Indiana and Florida were residuals.  So, I ran a regression where I took those two states out and the results are as follows:

Multiple R                           0.982
R Square                              0.964
Adjusted R Square              0.963

PDI t-stat                             34.57

A Multiple R value of .982!! 

The R Square and Adjusted R Square values are extremely close with .964 and .963 respectively.  The t-stat value is 34.57 – almost twice that of the previous regression with Indiana and Florida.

Residual analysis:
Indiana – the state of Indiana has half the population of Ohio, almost same distance and similar adjusted median incomes and yet Indiana had slightly more sightings than Ohio.  This would cause the regression to categorize this state as a residual.

Florida – the state of Florida is listed with a distance of 1,438 miles away to the largest metro area, Miami, and yet the state comes in 4th place for number of observations at 88 observations.  Even Texas which came in 5th place with 38 sightings – Texas has a larger population and closer and had less than half the observations that Florida had.  So, the snowbirds were probably influencing the number of Florida license plates that were seen.

Summary
The findings from this study are extremely similar to the findings from the Summer report.  The population variable in this study is showing as being more significant than it was in the Summer report.  The income variable continues to remain significant as well.



Sources –

















Appendix 1 – Data Set
State
Observations
Population
Distance
PCIA
Illinois
118
12,859,995
201
$37,223
Ohio
96
11,613,423
234
$37,080
Indiana
67
6,619,680
251
$36,164
Florida
66
20,271,272
1,432
$31,652
Texas
42
27,469,114
1,135
$37,311
New York
33
19,795,791
677
$28,948
Wisconsin
28
5,771,337
315
$33,853
Georgia
26
10,214,860
768
$34,476
California
23
39,144,818
2,230
$28,651
Tennessee
22
6,600,299
543
$35,238
Kentucky
21
4,425,092
367
$34,054
Pennsylvania
20
12,802,503
657
$34,020
North Carolina
17
10,042,802
679
$32,564
Virginia
17
8,382,993
792
$38,351
Minnesota
17
5,489,594
631
$36,017
Massachusetts
12
6,794,422
786
$33,171
Colorado
12
5,456,574
1,190
$36,314
Arizona
12
6,828,065
1,939
$34,179
New Jersey
11
8,958,013
669
$32,269
Maryland
11
6,006,401
582
$32,968
Missouri
9
6,083,672
482
$34,538
Alabama
9
4,858,979
754
$33,273
Connecticut
9
3,590,886
717
$31,744
Arkansas
9
2,978,204
836
$30,767
Iowa
8
3,123,899
537
$34,739
Louisiana
8
4,670,724
1,065
$33,138
South Carolina
7
4,896,146
715
$30,345
Oklahoma
6
3,911,338
972
$34,556
Washington
6
7,170,351
2,269
$39,802
Nebraska
6
1,896,190
678
$35,075
Oregon
6
4,028,977
2,328
$26,238
Maine
5
1,329,328
919
$28,393
Mississippi
4
2,992,333
927
$33,174
Kansas
3
2,911,641
671
$35,682
Nevada
3
2,890,845
1,885
$31,643
New Mexico
3
2,085,109
1,490
$31,016
South Dakota
3
858,469
797
$30,029
Montana
3
1,032,949
1,681
$29,522
Alaska
3
738,432
3,792
$34,772
Idaho
2
1,654,930
1,911
$33,734
Vermont
2
626,042
721
$28,857
Wyoming
2
586,107
1,168
$41,250
Utah
1
2,995,919
1,608
$34,946
New Hampshire
1
1,330,608
812
$31,540
Delaware
1
945,934
644
$35,060
Rhode Island
1
1,056,298
774
$30,483
North Dakota
1
756,927
865
$37,296
West Virginia
0
1,844,128
387
$28,066


Appendix 1 – Data Set
The states on the left are color coded based on their initial predicted difficulty of finding that particular state in Michigan.
Dark Green – Very easy to find
Light Green – Easy to find
Yellow – Moderate
Light Blue – Difficult to find
Dark Blue – Very difficult to find


















Appendix 2 – Week by Week Multiple Regressions
Population, Distance, and PCIA - 48 States
Week #
Multiple R
R Squared
Adjusted R Squared
Pop t-stat
Distance t-stat
PCIA t-stat
Week 1
0.68999
0.47609
0.44037
5.67576
-2.27366
1.95812
Week 1 to 2
0.67033
0.44935
0.41180
5.46024
-1.92134
1.84606
Week 1 to 3
0.67939
0.46157
0.42486
5.37286
-2.52214
1.91417
Week 1 to 4
0.66125
0.43725
0.39888
5.16111
-2.26594
1.85274
Week 1 to 5
0.65906
0.43435
0.39579
4.93319
-2.48533
2.06433
Week 1 to 6
0.64544
0.41659
0.37681
4.75895
-2.41012
1.97090
Week 1 to 7
0.63701
0.40578
0.36527
4.65614
-2.36140
1.92011
Week 1 to 8
0.64487
0.41585
0.37602
4.78044
-2.42528
1.84001
Week 1 to 9
0.64995
0.42243
0.38305
4.84050
-2.53545
1.78507
Week 1 to 10
0.63939
0.40882
0.36852
4.59768
-2.57547
1.84526
Week 1 to 11
0.64953
0.42189
0.38247
4.71575
-2.66636
1.88675
Week 1 to 12
0.62235
0.38731
0.34554
4.23044
-2.57122
1.96579
Week 1 to 13
0.62797
0.39434
0.35305
4.29445
-2.58309
2.02354
Week 1 to 14
0.62702
0.39316
0.35178
4.28894
-2.56872
2.02010
Week 1 to 15
0.63238
0.39990
0.35898
4.36887
-2.57404
2.04862
Week 1 to 16
0.63466
0.40279
0.36207
4.39208
-2.56959
2.08604
Week 1 to 17
0.64664
0.41814
0.37847
4.55733
-2.66650
2.09562
Population and Distance - 48 States
Multiple R
R Squared
Adjusted R Squared
Pop t-stat
Distance t-stat
Week 1
0.65607
0.43043
0.40512
5.47722
-2.36921
Week 1 to 2
0.63773
0.40670
0.38033
5.29341
-2.02957
Week 1 to 3
0.64555
0.41674
0.39081
5.19346
-2.61804
Week 1 to 4
0.62718
0.39335
0.36639
5.00058
-2.36828
Week 1 to 5
0.61609
0.37957
0.35200
4.73427
-2.57466
Week 1 to 6
0.60422
0.36508
0.33686
4.58534
-2.50474
Week 1 to 7
0.59665
0.35599
0.32737
4.49553
-2.45599
Week 1 to 8
0.60902
0.37090
0.34294
4.63216
-2.53486
Week 1 to 9
0.61693
0.38061
0.35308
4.70141
-2.64288
Week 1 to 10
0.60256
0.36308
0.33477
4.45308
-2.68279
Week 1 to 11
0.61247
0.37512
0.34735
4.56006
-2.77042
Week 1 to 12
0.57750
0.33350
0.30388
4.07548
-2.69390
Week 1 to 13
0.58136
0.33798
0.30855
4.12681
-2.70414
Week 1 to 14
0.58041
0.33687
0.30740
4.12190
-2.68754
Week 1 to 15
0.58537
0.34266
0.31344
4.19374
-2.69176
Week 1 to 16
0.58628
0.34373
0.31456
4.20911
-2.68970
Week 1 to 17
0.60006
0.36007
0.33162
4.36644
-2.78088



Any R Squared or Adjusted R Squared value above .3 is highlighted in light green. 
Any R Squared value or Adjusted R Squared value above .5 is highlighted in dark green.
t-stat values above 2.0 are highlighted in yellow.



Appendix 2 – Week by Week Multiple Regressions
Population, Distance, and PCIA - 42 States
Multiple R
R Squared
Adjusted R Squared
Pop t-stat
Distance t-stat
PCIA t-stat
Week 1
0.68718
0.47222
0.43055
5.79623
-1.73763
1.07487
Week 1 to 2
0.73197
0.53578
0.49913
6.56111
-1.67403
1.43147
Week 1 to 3
0.72894
0.53136
0.49436
6.43210
-2.58248
1.47521
Week 1 to 4
0.77621
0.60250
0.57112
7.50473
-2.72679
1.42596
Week 1 to 5
0.78942
0.62318
0.59343
7.80785
-2.95929
1.66834
Week 1 to 6
0.81361
0.66197
0.63528
8.47994
-3.43116
1.69654
Week 1 to 7
0.79663
0.63462
0.60578
7.95959
-3.30987
1.71070
Week 1 to 8
0.82210
0.67585
0.65026
8.78506
-3.41517
1.35390
Week 1 to 9
0.83261
0.69324
0.66902
9.12391
-3.75422
1.20163
Week 1 to 10
0.85143
0.72493
0.70321
9.84039
-4.10730
1.32752
Week 1 to 11
0.84286
0.71041
0.68755
9.44856
-4.13941
1.45965
Week 1 to 12
0.83715
0.70082
0.67721
9.26433
-3.85128
1.60339
Week 1 to 13
0.84240
0.70963
0.68671
9.43714
-3.98507
1.79387
Week 1 to 14
0.84546
0.71480
0.69228
9.52883
-4.11162
1.93786
Week 1 to 15
0.85498
0.73099
0.70975
9.92849
-4.20511
2.08147
Week 1 to 16
0.85164
0.72530
0.70361
9.75148
-4.16976
2.27066
Week 1 to 17
0.84617
0.71601
0.69359
9.50113
-4.23776
2.19424
Population and Distance - 42 States
Multiple R
R Square
Adjusted R Squared
Pop t-stat
Distance t-stat
Week 1
0.67540
0.45617
0.42828
5.71555
-1.62652
Week 1 to 2
0.71466
0.51074
0.48565
6.38034
-1.50978
Week 1 to 3
0.71029
0.50452
0.47911
6.23903
-2.40276
Week 1 to 4
0.76239
0.58123
0.55976
7.31647
-2.55536
Week 1 to 5
0.77174
0.59558
0.57484
7.52600
-2.73155
Week 1 to 6
0.79772
0.63636
0.61772
8.17377
-3.18926
Week 1 to 7
0.77877
0.60649
0.58630
7.65796
-3.06567
Week 1 to 8
0.81254
0.66022
0.64279
8.61424
-3.26624
Week 1 to 9
0.82558
0.68158
0.66525
9.00649
-3.63555
Week 1 to 10
0.84390
0.71217
0.69741
9.67253
-3.96155
Week 1 to 11
0.83317
0.69417
0.67849
9.23036
-3.96253
Week 1 to 12
0.82498
0.68058
0.66420
8.99015
-3.65925
Week 1 to 13
0.82767
0.68504
0.66889
9.07379
-3.74530
Week 1 to 14
0.82862
0.68661
0.67054
9.09295
-3.82913
Week 1 to 15
0.83685
0.70032
0.68495
9.40487
-3.88083
Week 1 to 16
0.82947
0.68803
0.67203
9.13271
-3.79754
Week 1 to 17
0.82464
0.68003
0.66362
8.93431
-3.88058
Any R Squared or Adjusted R Squared value above .3 is highlighted in light green. 
Any R Squared value or Adjusted R Squared value above .5 is highlighted in dark green.

t-stat values above 2.0 are highlighted in yellow. 

No comments:

Post a Comment

Note: Only a member of this blog may post a comment.