Break Points - R-help | R Mailing Lists

Fri, Dec 30, 2011 3:14 AM #

Respected Sir

I tried the strucchange
My data is attached. However I tried the attached commands (last
save.txt) to perform Bai Perron 2003... I t worked well but in the end
it is giving warning that overlapping confidence interval... I am not
sure how to proceed... Please Help Me
Thanking You
Ayanendu Sanyal

Please have a look at our new mission and contribute into it (cut and
paste the link below in the address bar of your internet browser)

http://thesocialscienceinformer.blogspot.com/

Thanking you

Ayanendu Sanyal
PhD Scholar
Institute for Social and Economic Change (ISEC)
P.O- Nagarbhavi
Bangalore-72
State- Karnataka
Country- India
PIN- 560072

www.isec.ac.in/phd.html

http://ayanendusanyal.blogspot.com/
-------------- next part --------------
t	lnrpe
1	1.6113515
2	1.619601724
3	1.599889264
4	1.645954835
5	1.723777317
6	1.830606002
7	2.034751407
8	2.112377045
9	2.095050993
10	2.046822835
11	2.276628064
12	2.543864584
13	2.619425807
14	2.717786454
15	2.874537082
16	2.923223972
17	3.136825311
18	3.206377996
19	3.352655132
20	3.49032806
21	3.508602739
22	3.621768106
23	3.803617305
24	4.141727497
25	4.27471221
26	4.34523451
27	4.242555261
28	4.262046942
29	4.378894917
30	4.419018243
31	4.391496862
32	4.489015146
33	4.588180885
34	4.846861554
35	5.069645376
36	5.257766481
37	5.292695491
38	5.307844982
39	5.277006289
40	5.323613873
41	5.377629069
42	5.429256311
43	5.443411354
44	5.47242567
45	5.727392687
46	6.147054773
-------------- next part --------------
> ayan <- read.table("ayan.txt", header=T)
> x =read.table(file="ayan.txt",header=T)
> library(strucchange)
> bp.ayan <- breakpoints(x[["lnrpe"]] ~ x[["t"]])
> summary (bp.ayan)

         Optimal (m+1)-segment partition: 

Call:
breakpoints.formula(formula = x[["lnrpe"]] ~ x[["t"]])

Breakpoints at observation number:
                         
m = 1         23         
m = 2         23    33   
m = 3   11    23    33   
m = 4   11    23    33 40
m = 5   11    20 26 34 40
m = 6   8  14 20 26 34 40

Corresponding to breakdates:
                                                                               
m = 1                                       0.5                                
m = 2                                       0.5                                
m = 3   0.239130434782609                   0.5                                
m = 4   0.239130434782609                   0.5                                
m = 5   0.239130434782609                   0.434782608695652 0.565217391304348
m = 6   0.173913043478261 0.304347826086956 0.434782608695652 0.565217391304348
                                           
m = 1                                      
m = 2   0.717391304347826                  
m = 3   0.717391304347826                  
m = 4   0.717391304347826 0.869565217391304
m = 5   0.739130434782609 0.869565217391304
m = 6   0.739130434782609 0.869565217391304

Fit:
                                                                           
m   0           1           2           3           4           5          
RSS   1.2131579   0.7466773   0.5243361   0.3570253   0.2712234   0.2589809
BIC -25.2008012 -36.0409283 -40.8160181 -47.0090984 -48.1669059 -38.8056613
               
m   6          
RSS   0.2698858
BIC -25.4224766
> plot (bp.ayan)
> breakdates(bp.ayan)
[1] 0.2391304 0.5000000 0.7173913 0.8695652
> ci.ayan <- confint(bp.ayan)
> breakdates(ci.ayan)
      2.5 % breakpoints    97.5 %
1 0.2173913   0.2391304 0.3260870
2 0.4782609   0.5000000 0.5217391
3 0.6956522   0.7173913 0.7391304
4 0.5434783   0.8695652 0.8913043
Warning message:
Overlapping confidence intervals 
> ci.ayan

         Confidence intervals for breakpoints
         of optimal 5-segment partition: 

Call:
confint.breakpointsfull(object = bp.ayan)

Breakpoints at observation number:
  2.5 % breakpoints 97.5 %
1    10          11     15
2    22          23     24
3    32          33     34
4    25          40     41

Corresponding to breakdates:
      2.5 % breakpoints    97.5 %
1 0.2173913   0.2391304 0.3260870
2 0.4782609   0.5000000 0.5217391
3 0.6956522   0.7173913 0.7391304
4 0.5434783   0.8695652 0.8913043
Warning message:
Overlapping confidence intervals

Achim Zeileis

Fri, Dec 30, 2011 1:15 PM #

On Fri, 30 Dec 2011, Ayanendu Sanyal wrote:

Overlapping confidence intervals typically indicate that too many 
breakpoints were selected and/or the model does not fit well. In your 
case, the BIC selection is probably not the best solution.

hth,
Z

Ayanendu Sanyal

Sat, Dec 31, 2011 7:09 AM #

Respected Sir
Thank You for the ultra fast reply.... I changed the file and defined
my data as time series and then ran 'strucchange'.... i have attached
the lastsave1.txt file..... but the lines (ci.ayan) is not working....
Is the work ok???? please confirm .... I am not so much familar with
this test... This time it didnt gave any warning message.
Ayan

On Sat, Dec 31, 2011 at 2:45 AM, Achim Zeileis <Achim.Zeileis at uibk.ac.at> wrote:

Please have a look at our new mission and contribute into it (cut and
paste the link below in the address bar of your internet browser)

http://thesocialscienceinformer.blogspot.com/

Thanking you

Ayanendu Sanyal
PhD Scholar
Institute for Social and Economic Change (ISEC)
P.O- Nagarbhavi
Bangalore-72
State- Karnataka
Country- India
PIN- 560072

www.isec.ac.in/phd.html

http://ayanendusanyal.blogspot.com/
-------------- next part --------------

R version 2.14.1 (2011-12-22)
Copyright (C) 2011 The R Foundation for Statistical Computing
ISBN 3-900051-07-0
Platform: i386-pc-mingw32/i386 (32-bit)

R is free software and comes with ABSOLUTELY NO WARRANTY.
You are welcome to redistribute it under certain conditions.
Type 'license()' or 'licence()' for distribution details.

  Natural language support but running in an English locale

R is a collaborative project with many contributors.
Type 'contributors()' for more information and
'citation()' on how to cite R or R packages in publications.

Type 'demo()' for some demos, 'help()' for on-line help, or
'help.start()' for an HTML browser interface to help.
Type 'q()' to quit R.

> library(strucchange)
Loading required package: zoo

Attaching package: ?zoo?

The following object(s) are masked from ?package:base?:

    as.Date, as.Date.numeric

Loading required package: sandwich
> ayan <- read.table("ayan.txt", header=T)
> ayan
      lnrpe
1  1.611352
2  1.619602
3  1.599889
4  1.645955
5  1.723777
6  1.830606
7  2.034751
8  2.112377
9  2.095051
10 2.046823
11 2.276628
12 2.543865
13 2.619426
14 2.717786
15 2.874537
16 2.923224
17 3.136825
18 3.206378
19 3.352655
20 3.490328
21 3.508603
22 3.621768
23 3.803617
24 4.141727
25 4.274712
26 4.345235
27 4.242555
28 4.262047
29 4.378895
30 4.419018
31 4.391497
32 4.489015
33 4.588181
34 4.846862
35 5.069645
36 5.257766
37 5.292695
38 5.307845
39 5.277006
40 5.323614
41 5.377629
42 5.429256
43 5.443411
44 5.472426
45 5.727393
46 6.147055
> ayan= ts(ayan, start=1.611352)
> ayan
Time Series:
Start = 1.611352 
End = 46.61135 
Frequency = 1 
         lnrpe
 [1,] 1.611352
 [2,] 1.619602
 [3,] 1.599889
 [4,] 1.645955
 [5,] 1.723777
 [6,] 1.830606
 [7,] 2.034751
 [8,] 2.112377
 [9,] 2.095051
[10,] 2.046823
[11,] 2.276628
[12,] 2.543865
[13,] 2.619426
[14,] 2.717786
[15,] 2.874537
[16,] 2.923224
[17,] 3.136825
[18,] 3.206378
[19,] 3.352655
[20,] 3.490328
[21,] 3.508603
[22,] 3.621768
[23,] 3.803617
[24,] 4.141727
[25,] 4.274712
[26,] 4.345235
[27,] 4.242555
[28,] 4.262047
[29,] 4.378895
[30,] 4.419018
[31,] 4.391497
[32,] 4.489015
[33,] 4.588181
[34,] 4.846862
[35,] 5.069645
[36,] 5.257766
[37,] 5.292695
[38,] 5.307845
[39,] 5.277006
[40,] 5.323614
[41,] 5.377629
[42,] 5.429256
[43,] 5.443411
[44,] 5.472426
[45,] 5.727393
[46,] 6.147055
> bp.ayan <- breakpoints(ayan ~ 1)
> summary (bp.ayan)

         Optimal (m+1)-segment partition: 

Call:
breakpoints.formula(formula = ayan ~ 1)

Breakpoints at observation number:
                        
m = 1           22      
m = 2        16    33   
m = 3     11    22 34   
m = 4     10 16 23 34   
m = 5     10 16 23 33 40
m = 6   6 12 18 24 33 40

Corresponding to breakdates:
                                                                  
m = 1                                22.611352                    
m = 2                      16.611352           33.611352          
m = 3            11.611352           22.611352 34.611352          
m = 4            10.611352 16.611352 23.611352 34.611352          
m = 5            10.611352 16.611352 23.611352 33.611352 40.611352
m = 6   6.611352 12.611352 18.611352 24.611352 33.611352 40.611352

Fit:
                                                                     
m   0          1          2          3          4          5         
RSS  85.184802  18.716087   8.291525   3.576609   2.244549   1.809104
BIC 166.543984 104.491043  74.697447  43.677063  29.902475  27.638800
              
m   6         
RSS   1.617595
BIC  30.149073
> ci.ayan <- confint(bp.ayan)
> ci.ayan

         Confidence intervals for breakpoints
         of optimal 6-segment partition: 

Call:
confint.breakpointsfull(object = bp.ayan)

Breakpoints at observation number:
  2.5 % breakpoints 97.5 %
1     9          10     11
2    15          16     17
3    22          23     24
4    32          33     34
5    34          40     42

Corresponding to breakdates:
      2.5 % breakpoints   97.5 %
1  9.611352    10.61135 11.61135
2 15.611352    16.61135 17.61135
3 22.611352    23.61135 24.61135
4 32.611352    33.61135 34.61135
5 34.611352    40.61135 42.61135
> lines(ci.ayan)
Error in int_abline(a = a, b = b, h = h, v = v, untf = untf, ...) : 
  plot.new has not been called yet
> plot (ayan)
> ayan <- read.table("ayan.txt", header=T)
> ayan= ts(ayan, start=1.611352)
> bp.ayan <- breakpoints(ayan ~ 1)
> summary (bp.ayan)

         Optimal (m+1)-segment partition: 

Call:
breakpoints.formula(formula = ayan ~ 1)

Breakpoints at observation number:
                        
m = 1           22      
m = 2        16    33   
m = 3     11    22 34   
m = 4     10 16 23 34   
m = 5     10 16 23 33 40
m = 6   6 12 18 24 33 40

Corresponding to breakdates:
                                                                  
m = 1                                22.611352                    
m = 2                      16.611352           33.611352          
m = 3            11.611352           22.611352 34.611352          
m = 4            10.611352 16.611352 23.611352 34.611352          
m = 5            10.611352 16.611352 23.611352 33.611352 40.611352
m = 6   6.611352 12.611352 18.611352 24.611352 33.611352 40.611352

Fit:
                                                                     
m   0          1          2          3          4          5         
RSS  85.184802  18.716087   8.291525   3.576609   2.244549   1.809104
BIC 166.543984 104.491043  74.697447  43.677063  29.902475  27.638800
              
m   6         
RSS   1.617595
BIC  30.149073
> ci.ayan <- confint(bp.ayan)
> ci.ayan

         Confidence intervals for breakpoints
         of optimal 6-segment partition: 

Call:
confint.breakpointsfull(object = bp.ayan)

Breakpoints at observation number:
  2.5 % breakpoints 97.5 %
1     9          10     11
2    15          16     17
3    22          23     24
4    32          33     34
5    34          40     42

Corresponding to breakdates:
      2.5 % breakpoints   97.5 %
1  9.611352    10.61135 11.61135
2 15.611352    16.61135 17.61135
3 22.611352    23.61135 24.61135
4 32.611352    33.61135 34.61135
5 34.611352    40.61135 42.61135
> 
-------------- next part --------------
lnrpe
1.6113515
1.619601724
1.599889264
1.645954835
1.723777317
1.830606002
2.034751407
2.112377045
2.095050993
2.046822835
2.276628064
2.543864584
2.619425807
2.717786454
2.874537082
2.923223972
3.136825311
3.206377996
3.352655132
3.49032806
3.508602739
3.621768106
3.803617305
4.141727497
4.27471221
4.34523451
4.242555261
4.262046942
4.378894917
4.419018243
4.391496862
4.489015146
4.588180885
4.846861554
5.069645376
5.257766481
5.292695491
5.307844982
5.277006289
5.323613873
5.377629069
5.429256311
5.443411354
5.47242567
5.727392687
6.147054773

Achim Zeileis

Sat, Dec 31, 2011 8:46 AM #

On Sat, 31 Dec 2011, Ayanendu Sanyal wrote:

Now you fit a different model to your data (just a constant) which does 
not seem to be very reasonable. You should use a regression model that 
makes sense for the data - at least within every segment.
hth,
Z