Table of Contents Table of Contents
Previous Page  174 / 322 Next Page
Information
Show Menu
Previous Page 174 / 322 Next Page
Page Background

期貨未平倉量的資訊內涵及其交易活動之研究

174

much threat to our results because we have large sample size (2,226 observations) relatively

to the number of parameters (10 parameters,

p

= 1, 3, 19, 20, 21 and

q

= 3, 24, 26, 39, 44) to

be estimated. In our case, interpretation of the decomposed series is unimportant because the

purpose of model is to separate the expected and unexpected components of

OI

series. The

model describes the “expected”

OI

series, whose time series property has never been

discussed in any existing theory or paper. As a result, interpreting the contents and properties

of the decomposed series is less in need for this case (see similar argument in Bessembinder

and Seguin (1993)).

It appears that the advantages of the more precise model outweigh its disadvantages.

For these reasons, we favor the more precise model over the parsimonious ones.

To be concrete, we test the robustness of our main findings using alternative models

with more parsimonious form. We use ARIMA Models (1) to (4) in Table 7 as alternative

models for the decomposition of

OI

. We then used the decomposed series from each

alternative model to perform regressions Model (2), (3), and (4) in Table 3 and 4. The results

of the final model are unaltered except in few occasions.

10

The tests indicate that the

empirical results presented in the manuscript are robust to different models of OI

decomposition.

Table 7 Alternative ARIMA Models for Robustness Analysis

Model

p

(AR terms)

q

(MA terms)

AIC

SBC ACF & PACF Q-statistics

Alternative Models

1

1,2

1

19.37242 19.38012 20,24,44 20(0.000)

2

1,2,3,4,5,6 1,2,3

19.37088 19.39401 20,24,44 20(0.000)

3

1~19

1,2,3

19.34881 19.40563 23,44

44(0.000)

4

1,2,3,19

1,2,3

19.34635 19.36185 20,44

20(0.000)

5

1,3,19,20,21 3,20,24,26,39,44 19.30113 19.32956 40

12(0.085)

13(0.083)

60(0.073)

6

1,3,19,20,21 20,24,26,39,44 19.30276 19.32860 -

11,12…

(>0.020)

7

1,3,20,21 3,20,24,26,39,44 19.30231 19.32815 -

56(0.074)

60(0.03)

Final Model

8

1,3,19,20,21 3,24,26,39,44 19.31195 19.33779 -

-

10 Results for this robustness test are not reported for the purpose of brevity. A summary table of the findings

using alternative decomposition models is available upon request.