rerun with adf test on mobility

After filtering moblity on adf p-value, the number available mobility test are much smaller. 

On ECS323 Lambda server 
K02, 
18:28, auto_county_cajo_20200501-20210215-01percent-K02.R
18:33 auto_county_cajo_20200501-20210215-05percent-K02.R
18:39 auto_county_cajo_20200501-20210215-10percent-K02.R
19:02 K02, summary and plot udpated. Similar figures. 

K03
19:20.  auto_county_cajo_20200501-20210215-01percent-K03.R
19:27 auto_county_cajo_20200501-20210215-05percent-K03.R

19:32, auto_county_cajo_20200501-20210215-10percent-K03.R

K04

19:42 auto_county_cajo_20200501-20210215-01percent-K04.R

19:47 auto_county_cajo_20200501-20210215-05percent-K04.R

19:52 auto_county_cajo_20200501-20210215-10percent-K04.R

K05

20:15 auto_county_cajo_20200501-20210215-01percent-K05.R

20:32 auto_county_cajo_20200501-20210215-05percent-K05.R

20:44 auto_county_cajo_20200501-20210215-10percent-K05.R

K06

20:56 auto_county_cajo_20200501-20210215-01percent-K06.R

21:05 auto_county_cajo_20200501-20210215-05percent-K06.R

21:11 auto_county_cajo_20200501-20210215-10percent-K06.R

21:42, Redo all plots, 

21:43 - 23:11, Rmd to generate summary table based  cval=1% cajo test. There was an index error to find out the appropriate numerators. It should be chr '2' or chr '3', not integer 2 or 3. This is a very tricky error. 

Counties with adf_p-value > 0.2 for dewpoint

McKenzie	North Dakota
Williams	North Dakota
Fresno	California
Flathead	Montana
Madera	California
Mountrail	North Dakota
Stark	North Dakota
Juneau	Alaska
Pennington	South Dakota
Santa Cruz	Arizona
Meade	South Dakota
Lewis and Clark	Montana
Morton	North Dakota
Cascade	Montana
Tulare	California
Ward	North Dakota
Keith	Nebraska
Pima	Arizona
Mariposa	California
Douglas	Washington
Burleigh	North Dakota
Lake	Montana
Stutsman	North Dakota

 

K05

13:33 10% run on lambda

13:40 5% cval run on lambda

13:49 1% cval run

plots done. 

K03

12:35 10% run on lamba

 12:42, 5% run

12:53, 1% run

K01

Chose 1% critical value for plot, since the results are so good. Too good to be true? 

 

K02

 

K=4 amazing results

 

K06 run

Rt-workplace good results

Rt - T2m poor result

 

K08

K08 show good cointegration of Rt - workplace mobility, but not Rt- T2m

 

K17

19:17 10% run

19:21, 5% run

19:33, 1% run

19:50, generate table and plot

K15 run

K14 - K16 seem to have good co-integrated Rt-T2m.  

K20 run

K20, Rt-T2m trends weakens. It seems to peak at K14-16. So, I should try K15.

17:00 lambda 10% run, 17:10 done

17:23, 5% run done. 

17:31, 01% run. done

generate table and plots. Add limits=c(0, 0.6) to plots. 

K18 run

 16:00 lamba, 5% R code run. HQ could not find the output file. 

16:25, 10% run on lambda. 

K16 conintegration

It seems K=16 days is the optimal lag between Rt and T2m. 

 

K14 cointegration

auto_county_cajo_20200501-20210215-01percent-K14.R 

15 month time window, no Rt-T2m rank of 2

 

I finished another cointegration analysis for 15 month year time window, 2020 May 1 - 2021 August 15. No rank of two is found for cajo Rt ~ T2m. So, either vaccination really alter the correlation, or second winter has to be included. 

cajo Rt - T2m pattern

 After filter adf p-value, NE cluster disappeared. Instead, Alaska and west coat are the new cluster. 

tb <- tb %>% filter( adf_p_Rt>0.01 & adf_p_t2m >0.05)

2020May1-Sep1 covid19 peak analysis

 pwd: 

(base) hqin@CS313BQin _run-2020501-20200901 % pwd

/Users/hqin/github/COVID19_transmission_MS/_run-2020501-20200901

Rt ~ T2m is very strong!!!

In the summer, T2m are mostly non-stationary!!!

cajo, Carter, TN,

 Cater, TN

time window April 1 to September 1, 2020. 

adf.test(Rt) --> p-value <=0.01. !!!!

This means Rt is stationary and leads to numerical errors in further analysis. 

two independent random walks, r=0 for Johansen test

 two independent random walks, r=0 for Johansen test

one random walk and one stationary series, r=1 for Johansen test.