Plot innovation variance and covariance parameters*

[Mario Ferri]

I run two types of models: a rolling VAR and a rolling PVAR, and I need to run both.

1)
collect the observation error covariance parameters and plot the credible intervals

And

2) Collect the (state) innovation variance parameters and plot the credible intervals.

How do I do that?

Data sample for the standard VAR are given below The code I run are:

 tsset year 
quietly rolling, window(12) clear : var mag0 mag1 car0 car1 col0 col1 vehicle ka re fd , lags(1/1)

and :

quietly rolling, window(12) clear : pvar mag0 mag1 car0 car1 col0 col1 vehicle ka re fd , lags(1/1)

* Example generated by -dataex-. For more info, type help dataex
clear
input float(year mag0 mag1 car0 car1 ka vehicle col0 col1 re fd)
1945 .08333334 .9166667        1           0           .            .            .             .             .             .
1946 .11104685 .8889531 .9972528  .002747253           .     .0767169   .007133216   -.007133186             .             .
1947 .19947444 .8005255 .9980889 .0019111324           .  -.033678144    .05756658    -.05756658             .             .
1948 .22086407 .7406744 .9615384           0           .    .02165809  -.034264237  -.0041972995             .             .
1949  .1937322 .7993488  .993081           0           .  -.003348023   .000801742      .0307408             .             .
1950  .2035002 .7964998  .992267  .007733008           .  .0028122663  -.030423284     .03734231             .             .
1951 .22954823 .7704518 .9502442    .0497558           .  -.003260791    .05352053    -.05352056             .             .
1952 .16732623 .8326738 .9955353 .0044647385           .   .033309698    -.0533452      .0533452             .             .
1953 .16870166 .8312984 .9815832  .018416768           .  -.008517623    .08407372    -.08407375             .             .
1954 .17623118 .8237689 .9855515  .014448515           .  -.008653492  -.013736248    .013736278             .             .
1955  .1921042 .8078958 .9871795  .012820513           .   .005571097    -.0450753      .0450753             .             .
1956 .18711314 .8128868 .9950279  .004972095           .  -.013731807  -.008823812    .008823812             .             .
1957  .1980057 .7987383  .996744           0           .   .006495297     .0465731    -.04982913             .             .
1958 .17531542 .8002645 .9734432 .0021367522           .  .0042423904    .03551078    -.05667481             .             .
1959  .3041311  .695869 .9708995   .02910053           . -.0012779236    .09544164    -.07102156             .             .
1960 .26169458 .7383054        1           0           .  -.015391976   -.03749508     .03749505             .             .
1961 .27614978 .7238503 .9675417  .032458283           .    .00309664    .03749508    -.03749505             .             .
1962 .27188328 .7281167 .9976317 .0023683214           .   .010390967   -.11199883     .11199877             .             .
1963  .3119553 .6880447  .969496   .03050398           .   -.01049766  -.001894653    .001894653             .             .
1964  .3092237 .6907762 .9722375   .02776256           .   .004954934   .015789598   -.015789539             .             .
1965  .2330586 .7669414        1           0           .     .0115287   -.05335116     .05335113             .             .
1966 .17653315 .8234668 .9927331  .007266927           .   .012013525    .02140197      -.021402             .             .
1967 .22536334 .7746366 .9595888    .0404112           .  -.011647135    .04271534    -.04271531             .             .
1968  .1899249 .8100751 .9677419  .032258064           .    .01712513  -.018935263    .018935263             .             .
1969 .19505495 .8049451 .9677419  .032258064           .  -.014389515  -.007119328    .007119298             .             .
1970  .2045374 .7899681 .9622474  .032258064           .  .0019976795   .025434285    -.03092873             .             .
1971  .2260723 .7739277 .9621588   .03784119  -.39362165  -.021190494   .006380677  -.0008862019             .             .
1972 .27264693 .7169244 .9406098   .04896155  -.24231456  -.007597685    .05811477   -.068543434             .             .
1973 .21797733 .7507727 .9263393   .04241072  -.24066724   .006259322  -.021107137  .00028583407             .             .
1974 .20072116 .7578125 .9370707   .02146291   -.2665547   .005590796    .02026099   -.030477345             .             .
1975 .17659684 .8094093 .9783654  .007640797  -.24634326   .006330937    .01244849    .015024036             .             .
1976   .184973  .815027 .9853051  .014694895  -.31063345    .01055935 -.0037447214    .017738491             .             .
1977 .23524334 .7647567 .9867347  .013265306   -.4182096 -.0006091893    .05594742    -.05594736             .             .
1978 .21381475 .7764521 .9846154  .005651491   -.3896763 -.0009466112  -.018916786    .009183645             .             .
1979 .28434065 .6985294 .9394796   .04339043   -.3896763    .01392892  .0014082193   -.008805037             .             .
1980  .1752619 .8247381 .9987913 .0012087027  -.29550555   -.02210301    .01566708   .0014628172             .             .
1981  .2135585 .7864415 .9974144   .00258565  -.34505615  -.010849535  -.015343845    .015343904             .             .
1982  .3413058 .6586943 .9786684   .02133161  -.25163704   .004226178    .02456367    -.02456367   .0001324024    .006160483
1983 .26389787 .7361021 .9928895  .007110537   -.1750894   -.00328362   -.06221724     .06221718   -.014355402    .014019087
1984 .18436745 .8156326        1           0  -.10805306  .0010181367   -.05473828     .05473831   -.007924562    .014482632
1985  .1516645 .8483355        1           0  -.08923411 -.0006535053    .02088243    -.02088246    .005102764    .009850457
1986 .19957983 .8004202        1           0   -.0594031  .0020222664   .034017503   -.034017444   -.010541415     .02026184
1987 .23287007   .76713 .9941823  .005817712  -.14884342   .006771833     .0583387    -.05833873     .01484855     .01551783
1988 .22385173 .7761483 .9896051  .010394843   -.1206697  -.003155172    .01464212    -.01464212     .01575236      .0220432
1989 .23149644 .7685035 .9966871 .0033128636 -.036256615   .016436338   -.05698216     .05698213 -.00008889288  -.0046534836
1990 .27984852 .7201515  .994654  .005346005 .0005995822    .05697283  -.015793383    .015793443   -.018172212     .02081114
1991  .3284199 .6465201 .9749401           0   .07461239     43.27272    .02895367    -.05401364    .015567867    .008893073
1992  .3083437  .691594 .9999377           0   .12935932    .04439545   -.09506172      .1200594   -.018236343 -.00018396974
1993 .24354993 .7332776 .9604037  .016423794    .3613803    -1.770897   -.04099771    .017887473   -.005394331    .013688922
1994 .23602992 .7400631 .9677929   .00830021    .6830521     .0444107   -.09033784      .0896034    .010098682    .033943623
1995 .27747253 .7225274 .9669732  .033026755    .7524721   .022476196    .04459012    -.02068317      .0101888     .01792264
1996   .325723  .674277 .9697465  .030253503     .857617    .02706146   -.05218634     .05218637     .01386066   .0020535886
1997 .30558395 .6757603 .9518272  .029516995    .7424365    .01917267  -.036713243    .018057466   -.024909677     .01871976
1998 .29313186 .6818681 .9739011 .0010989011    .7895167    .01044464   -.04234445    .036000192     .00673125     .03958488
1999 .28042582 .7138736 .9756181   .01868132    .8327662   .000793457    .02678573   -.007486284    .008219708      .0237163
2000  .2688356 .7311644 .9853424  .014657534    .9176872   .029144287  -.016068995    .021769524    -.02601591   .0046441555
2001  .2554258 .7445742 .9921017  .007898351    .9703661   .031337738   .012772292   -.012772262    .007965174     .02297038
2002  .2105769  .789423  .982967  .017032966   1.0378017   .015750885   -.07252747     .07252747    .005473087   -.006904125
2003  .1955357 .8044643  .985989   .01401099   1.1644768   .016868591 .00020602345 -.00020605326    .018079296   -.007299632
2004  .2408904 .7591096        1           0    1.276395     .0190773   .012151897   -.012151897    .022712436     .01002726
2005  .2480769 .7519231        1           0   1.3937308    .01517105 -.0023991466   .0023991466   -.022985077    .015026867
2006  .2673077 .7326923 .9806319   .01936813    1.459883    .02935791   .030700564   -.030700505  -.0082911225    .011752129
2007 .26824456 .7317554 .9940829   .00591716   1.5610563   .024166107   -.03289485     .03289479     -.0102654    .023671627
2008  .2303533 .7696467        1           0   1.5881433    .02479172  -.013935775    .013935804    .019328535     .02137667
2009 .23055235 .7694477 .9828658  .017134182   1.5653305    .02153778    .06577581    -.06577581   -.010148779   -.015480518
2010 .29588655 .7041135 .9681467   .03185328   1.5544957   .015098572   -.02413714     .02413714   -.018703487      .0095371
2011  .3009355 .6958717 .9968072           0   1.5382434    .00982666   -.06095931     .05776656      .0073763   -.004389167
2012 .25005552 .7229174 .9703073 .0026656794    1.521991   .029712677  -.021239385   -.002594888     .01201999   -.008769155
2013 .20782596 .7834868  .972676   .01863677   1.4883434   .017562866  -.007940888     .02628064    -.02810171   -.009512424
2014 .22527473 .7747253 .9808435   .01915652   1.4627517    .02828598  -.028363526    .037050784    .018062308   -.000639975
2015         .        .        .           .    1.490982   .018558502            .             .   -.003878433  -.0006780028
2016         .        .        .           .    1.501817    2.8140755            .             .   -.029535156     .00385046
2017         .        .        .           .   1.5382434    -44.66236            .             .     .04417016  -.0021682382
2018         .        .        .           .   1.5207077            .            .             .    .002493127    .007712066
2019         .        .        .           .    1.483138            .            .             .   -.007930711   -.005730808
2020         .        .        .           .           .            .            .             .    -.01844586             .
end

[Maarten Buis]

I don't know about var and pvar, so I cannot help you with that. But I suspect that the problem is that rolling overwrites the data, so once you have run quietly rolling, window(12) clear : var mag0 mag1 car0 car1 col0 col1 vehicle ka re fd , lags(1/1) you no longer have the data in memory and quietly rolling, window(12) clear : pvar mag0 mag1 car0 car1 col0 col1 vehicle ka re fd , lags(1/1) won't run. Even if you get you command to run by opening the data again, you loose the result from the first command.

This sounds like a case for frames. Here is a sketch of what that would look like:

// have the data in a frame called data
frames reset
frame create data
frame change data
use data.dta // open the data, it is probably not called data.dta but you get the point
tsset year

// copy the data to a frame called var
frame copy data var, replace
frame change var
quietly rolling, window(12) clear : var mag0 mag1 car0 car1 col0 col1 vehicle ka re fd , lags(1/1)

//copy the data to a frame called pvar
frame copy data pvar, replace
frame change pvar
quietly rolling, window(12) clear : pvar mag0 mag1 car0 car1 col0 col1 vehicle ka re fd , lags(1/1)

// collect the variable(s) you need from the frame var into frame pvar
frlink 1:1 start , frame(var)
frget your_variable_of_choice, from(var)

// now you can create the graph with both parameters

[Mario Ferri]

I don't know about var and pvar, so I cannot help you with that. But I suspect that the problem is that rolling overwrites the data, so once you have run quietly rolling, window(12) clear : var mag0 mag1 car0 car1 col0 col1 vehicle ka re fd , lags(1/1) you no longer have the data in memory and quietly rolling, window(12) clear : pvar mag0 mag1 car0 car1 col0 col1 vehicle ka re fd , lags(1/1) won't run. Even if you get you command to run by opening the data again, you loose the result from the first command.

This sounds like a case for frames. Here is a sketch of what that would look like:

// have the data in a frame called data
frames reset
frame create data
frame change data
use data.dta // open the data, it is probably not called data.dta but you get the point
tsset year

// copy the data to a frame called var
frame copy data var, replace
frame change var
quietly rolling, window(12) clear : var mag0 mag1 car0 car1 col0 col1 vehicle ka re fd , lags(1/1)

//copy the data to a frame called pvar
frame copy data pvar, replace
frame change pvar
quietly rolling, window(12) clear : pvar mag0 mag1 car0 car1 col0 col1 vehicle ka re fd , lags(1/1)

// collect the variable(s) you need from the frame var into frame pvar
frlink 1:1 start , frame(var)
frget your_variable_of_choice, from(var)

// now you can create the graph with both parameters

Thank you warmly for your time and consideration. Indeed, rolling windows overwrite the data, creating a new dataset.

Perhaps it was me not being clear in #1. If so, I apologize for that!


I have two types of VARs to run separately from each other. One is a basic rolling windows vector autoregression VAR, the var command, and the other is a rolling windows panel data vector autoregression VAR, the pvar command. I provided the data for the first basic rolling window vector autoregression VAR in order not to spam the forum along with the codes for the two types of models, thinking eventually wrongly that I should be able to get what I want from there. Similarly, for the rolling Windows panel data vector autoregression var, use the pvar command. I can provide data for the rolling Windows panel data vector autoregression VAR with the pvar command, should that be necessary.


I didn't find any e() stored for the variance and covariance in the help file after the rolling windows, nor could I figure out how to derive the variance and covariance parameters and then graph them.

To keep it short and clear, what I would like is something like the following two graphs I found in a research paper, which explore the credible intervals of each of the covariance and variance parameters. Some authors have used Monte Carlo burns in order to get similar results. I am happy to accept that approach as long as I can get results.


Thank you again for any help you could provide

Best,

Mario

[Mario Ferri]

Thank you warmly for your time and consideration. Indeed, rolling windows overwrite the data, creating a new dataset.

Perhaps it was me not being clear in #1. If so, I apologize for that!


I have two types of VARs to run separately from each other. One is a basic rolling windows vector autoregression VAR, the var command, and the other is a rolling windows panel data vector autoregression VAR, the pvar command. I provided the data for the first basic rolling window vector autoregression VAR in order not to spam the forum along with the codes for the two types of models, thinking eventually wrongly that I should be able to get what I want from there. Similarly, for the rolling Windows panel data vector autoregression var, use the pvar command. I can provide data for the rolling Windows panel data vector autoregression VAR with the pvar command, should that be necessary.


I didn't find any e() stored for the variance and covariance in the help file after the rolling windows, nor could I figure out how to derive the variance and covariance parameters and then graph them.

To keep it short and clear, what I would like is something like the following two graphs I found in a research paper, which explore the credible intervals of each of the covariance and variance parameters. Some authors have used Monte Carlo burns in order to get similar results. I am happy to accept that approach as long as I can get results.


Thank you again for any help you could provide

Best,

Mario

[ATTACH=CONFIG]n1715681[/ATTACH]


[ATTACH=CONFIG]n1715682[/ATTACH]

I found the

 designplot

command written by Nick Cox, but did not get or manage to produce the desired graph. I would be grateful for any help you could provide.

Best,

Mario

[Nick Cox]

Thanks for the mention, but I can't see any connection whatsoever between designplot (Stata Journal) and what little I understand here.

[Mario Ferri]

For the record, I found out that this can be plotted with state space models. Some programming is needed, though!