459:. High-dimensional mixed frequency time series regressions involve certain data structures that once taken into account should improve the performance of unrestricted estimators in small samples. These structures are represented by groups covering lagged dependent variables and groups of lags for a single (high-frequency) covariate. To that end, the machine learning MIDAS approach exploits the sparse-group LASSO (sg-LASSO) regularization that accommodates conveniently such structures. The attractive feature of the sg-LASSO estimator is that it allows us to combine effectively the approximately sparse and dense signals.
790:
37:
A MIDAS regression is a direct forecasting tool which can relate future low-frequency data with current and lagged high-frequency indicators, and yield different forecasting models for each forecast horizon. It can flexibly deal with data sampled at different frequencies and provide a direct forecast
438:
when applied in the context of mixed frequency data. Bai, Ghysels and Wright (2013) examine the relationship between MIDAS regressions and Kalman filter state space models applied to mixed frequency data. In general, the latter involves a system of equations, whereas, in contrast, MIDAS regressions
28:
with several co-authors. There is now a substantial literature on MIDAS regressions and their applications, including
Ghysels, Santa-Clara and Valkanov (2006), Ghysels, Sinko and Valkanov, Andreou, Ghysels and Kourtellos (2010) and Andreou, Ghysels and Kourtellos (2013).
439:
involve a (reduced form) single equation. As a consequence, MIDAS regressions might be less efficient, but also less prone to specification errors. In cases where the MIDAS regression is only an approximation, the approximation errors tend to be small.
184:
429:
38:
of the low-frequency variable. It incorporates each individual high-frequency data in the regression, which solves the problems of losing potentially useful information and including mis-specification.
541:
Andreou, Elena & Eric
Ghysels & Andros Kourtellos "Should macroeconomic forecasters use daily financial data and how?", Journal of Business and Economic Statistics 31, 240-251.
311:
261:
241:
55:
831:
532:
Andreou, Elena & Eric
Ghysels & Andros Kourtellos "Regression Models with Mixed Sampling Frequencies", Journal of Econometrics, 158, 246-261.
324:
860:
700:"midasml: Estimation and Prediction Methods for High-Dimensional Mixed Frequency Time Series Data maintained by Jonas Striaukas"
824:
850:
865:
855:
817:
671:
452:
686:"midasr: Mixed Data Sampling Regression maintained by Virmantas Kvedaras and Vaidotas Zemlys-Balevicius"
266:
614:
565:
456:
42:
467:
Several software packages feature MIDAS regressions and related econometric methods. These include:
615:"Machine learning panel data regressions with heavy-tailed dependent data: Theory and application"
246:
622:
577:
208:
46:
179:{\displaystyle y_{t}=\beta _{0}+\beta _{1}B(L^{1/m};\theta )x_{t}^{(m)}+\varepsilon _{t}^{(m)},}
640:
595:
801:
699:
685:
632:
587:
448:
16:
Econometric models involving data sampled at different frequencies are of general interest.
508:
Predicting
Volatility: How to Get Most Out of Returns Data Sampled at Different Frequencies
770:
318:
844:
435:
314:
797:
25:
636:
591:
713:
21:
644:
599:
613:
Babii, Andrii; Ball, Ryan T.; Ghysels, Eric; Striaukas, Jonas (2022-07-26).
566:"Machine Learning Time Series Regressions With an Application to Nowcasting"
727:
789:
434:
The regression models can be viewed in some cases as substitutes for the
477:
midasml, R package for High-Dimensional Mixed
Frequency Time Series Data
747:
752:
732:
424:{\displaystyle B(L^{1/m};\theta )=\sum _{k=0}^{K}B(k;\theta )L^{k/m}}
661:, Journal of Computational and Graphical Statistics, 22(2), 231-245.
627:
582:
775:
657:
Simon, N., J. Friedman, T. Hastie, and R. Tibshirani (2013):
564:
Babii, Andrii; Ghysels, Eric; Striaukas, Jonas (2022-07-03).
506:
Ghysels, Eric, Pedro Santa-Clara and Rossen
Valkanov (2006)
519:
Ghysels, Eric and Arthur Sinko and Rossen
Valkanov (2006)
550:
Bai, Jennie and Eric
Ghysels and Jonathan Wright (2013)
805:
521:
MIDAS Regressions: Further
Results and New Directions
327:
269:
249:
211:
58:
423:
305:
255:
235:
178:
455:. The machine learning MIDAS regressions involve
672:"MIDAS Matlab Toolbox maintained by Hang Qian"
201:denotes the frequency – for instance if
825:
570:Journal of Business & Economic Statistics
8:
714:"EViews 9.5 MIDAS Forecasting Demonstration"
832:
818:
626:
581:
411:
407:
379:
368:
342:
338:
326:
284:
280:
268:
248:
221:
216:
210:
161:
156:
137:
132:
109:
105:
89:
76:
63:
57:
45:appearing at a higher frequency than the
552:State Space Models and MIDAS Regressions
313:is a lag distribution, for instance the
499:
510:, Journal of Econometrics, 131, 59-95
7:
786:
784:
41:A simple regression example has the
554:, Econometric Reviews, 32, 779–813.
804:. You can help Knowledge (XXG) by
443:Machine Learning MIDAS Regressions
306:{\displaystyle B(L^{1/m};\theta )}
14:
523:, Econometric Reviews, 26, 53-90.
788:
447:The MIDAS can also be used for
400:
388:
358:
331:
300:
273:
228:
222:
168:
162:
144:
138:
125:
98:
1:
637:10.1016/j.jeconom.2022.07.001
592:10.1080/07350015.2021.1899933
256:{\displaystyle \varepsilon }
451:time series and panel data
236:{\displaystyle x_{t}^{(4)}}
193:is the dependent variable,
18:Mixed-data sampling (MIDAS)
882:
783:
24:regression developed by
861:Statistical forecasting
619:Journal of Econometrics
263:is the disturbance and
800:-related article is a
425:
384:
307:
257:
237:
180:
426:
364:
308:
258:
243:is quarterly –
238:
181:
851:Econometric modeling
659:A sparse-group LASSO
471:MIDAS Matlab Toolbox
457:Legendre polynomials
325:
267:
247:
209:
56:
43:independent variable
728:"MIDAS Python code"
688:. 23 February 2021.
232:
172:
148:
866:Econometrics stubs
856:Time series models
421:
303:
253:
233:
212:
197:is the regressor,
176:
152:
128:
47:dependent variable
813:
812:
474:midasr, R package
463:Software packages
33:MIDAS Regressions
873:
834:
827:
820:
792:
785:
758:
757:
744:
738:
737:
724:
718:
717:
710:
704:
703:
702:. 29 April 2022.
696:
690:
689:
682:
676:
675:
668:
662:
655:
649:
648:
630:
610:
604:
603:
585:
576:(3): 1094–1106.
561:
555:
548:
542:
539:
533:
530:
524:
517:
511:
504:
449:machine learning
430:
428:
427:
422:
420:
419:
415:
383:
378:
351:
350:
346:
312:
310:
309:
304:
293:
292:
288:
262:
260:
259:
254:
242:
240:
239:
234:
231:
220:
185:
183:
182:
177:
171:
160:
147:
136:
118:
117:
113:
94:
93:
81:
80:
68:
67:
881:
880:
876:
875:
874:
872:
871:
870:
841:
840:
839:
838:
781:
771:Distributed lag
767:
762:
761:
746:
745:
741:
726:
725:
721:
712:
711:
707:
698:
697:
693:
684:
683:
679:
670:
669:
665:
656:
652:
612:
611:
607:
563:
562:
558:
549:
545:
540:
536:
531:
527:
518:
514:
505:
501:
496:
465:
445:
403:
334:
323:
322:
321:. For example
276:
265:
264:
245:
244:
207:
206:
101:
85:
72:
59:
54:
53:
35:
12:
11:
5:
879:
877:
869:
868:
863:
858:
853:
843:
842:
837:
836:
829:
822:
814:
811:
810:
793:
779:
778:
773:
766:
763:
760:
759:
739:
719:
705:
691:
677:
663:
650:
605:
556:
543:
534:
525:
512:
498:
497:
495:
492:
491:
490:
489:Stata,midasreg
487:
484:
481:
478:
475:
472:
464:
461:
444:
441:
418:
414:
410:
406:
402:
399:
396:
393:
390:
387:
382:
377:
374:
371:
367:
363:
360:
357:
354:
349:
345:
341:
337:
333:
330:
302:
299:
296:
291:
287:
283:
279:
275:
272:
252:
230:
227:
224:
219:
215:
187:
186:
175:
170:
167:
164:
159:
155:
151:
146:
143:
140:
135:
131:
127:
124:
121:
116:
112:
108:
104:
100:
97:
92:
88:
84:
79:
75:
71:
66:
62:
34:
31:
13:
10:
9:
6:
4:
3:
2:
878:
867:
864:
862:
859:
857:
854:
852:
849:
848:
846:
835:
830:
828:
823:
821:
816:
815:
809:
807:
803:
799:
794:
791:
787:
782:
777:
774:
772:
769:
768:
764:
755:
754:
749:
748:"MIDAS Julia"
743:
740:
735:
734:
729:
723:
720:
715:
709:
706:
701:
695:
692:
687:
681:
678:
673:
667:
664:
660:
654:
651:
646:
642:
638:
634:
629:
624:
620:
616:
609:
606:
601:
597:
593:
589:
584:
579:
575:
571:
567:
560:
557:
553:
547:
544:
538:
535:
529:
526:
522:
516:
513:
509:
503:
500:
493:
488:
485:
482:
479:
476:
473:
470:
469:
468:
462:
460:
458:
454:
450:
442:
440:
437:
436:Kalman filter
432:
416:
412:
408:
404:
397:
394:
391:
385:
380:
375:
372:
369:
365:
361:
355:
352:
347:
343:
339:
335:
328:
320:
316:
315:Beta function
297:
294:
289:
285:
281:
277:
270:
250:
225:
217:
213:
204:
200:
196:
192:
173:
165:
157:
153:
149:
141:
133:
129:
122:
119:
114:
110:
106:
102:
95:
90:
86:
82:
77:
73:
69:
64:
60:
52:
51:
50:
48:
44:
39:
32:
30:
27:
23:
19:
806:expanding it
798:Econometrics
795:
780:
751:
742:
731:
722:
708:
694:
680:
666:
658:
653:
618:
608:
573:
569:
559:
551:
546:
537:
528:
520:
515:
507:
502:
466:
446:
433:
202:
198:
194:
190:
188:
40:
36:
26:Eric Ghysels
17:
15:
22:econometric
845:Categories
628:2008.03600
621:: 105315.
583:2005.14057
494:References
453:nowcasting
205:is yearly
645:0304-4076
600:0735-0015
398:θ
366:∑
356:θ
319:Almon Lag
298:θ
251:ε
154:ε
123:θ
87:β
74:β
765:See also
317:or the
753:GitHub
733:GitHub
643:
598:
483:Python
480:EViews
189:where
20:is an
796:This
776:ARMAX
623:arXiv
578:arXiv
486:Julia
802:stub
641:ISSN
596:ISSN
633:doi
588:doi
847::
750:.
730:.
639:.
631:.
617:.
594:.
586:.
574:40
572:.
568:.
431:.
49::
833:e
826:t
819:v
808:.
756:.
736:.
716:.
674:.
647:.
635::
625::
602:.
590::
580::
417:m
413:/
409:k
405:L
401:)
395:;
392:k
389:(
386:B
381:K
376:0
373:=
370:k
362:=
359:)
353:;
348:m
344:/
340:1
336:L
332:(
329:B
301:)
295:;
290:m
286:/
282:1
278:L
274:(
271:B
229:)
226:4
223:(
218:t
214:x
203:y
199:m
195:x
191:y
174:,
169:)
166:m
163:(
158:t
150:+
145:)
142:m
139:(
134:t
130:x
126:)
120:;
115:m
111:/
107:1
103:L
99:(
96:B
91:1
83:+
78:0
70:=
65:t
61:y
Text is available under the Creative Commons Attribution-ShareAlike License. Additional terms may apply.