8/13/2019 kinhte2

1/14

4.4 Operationalization of variables

4.4.1 Dependent variable: Impact of innovation

The Innovation Survey from Statistics Norway will be used in order to provide a reliable measure of

innovation. In accordance with the Oslo manual, SSB defines innovative companies as companies that

has introduced a new or significantly improved service or product or implemented significantly

improved processes in the period. The operationaliation of innovation in this paper is based on a

question in the survey asking, What percentage of your turnover in 2008 originates from significantly

improved services or products? This provides a reliable and detailed measure on the innovative activity

in the surveyed companies. By using this measure, we are able to capture not only innovation activities

(e.g. like a count of patents would do), but also capture impact of the innovations, as we only measure

innovations that actually provide a revenue stream to the firm.

In 2008, the Innovation survey received response from 614 firms within the offshore industry. The

number of respondents answering the question relating to what percentage of turnover originates from

new innovations was 215.

4.4.2 Independent variable: Academic publications

This variable aims at measuring to what degree impact of innovation in firms benefit from being located

in close proximity to research institutions. Using a broad definition, there are educational institutions

providing relevant education for the offshore industry in 18 of 19 counties in Norway (Oljeindustriens

Landsforening 2008). This includes all relevant educations equal to and higher than bachelor level.

However, following Rothaermel & Ku's (2008) reasoning, an educational institution (e.g. a University

College) is not sufficient for intellectual capital to benefit innovation. They argue that research needs to

be undertaken; hence there is an important difference between an educational institution and aresearch institution.

The relevant research institutions are selected by a review of published articles within relevant fields in

academic journals. For a detailed summary on this process, see Appendix 2. After manual corrections

(merging equal research institutions and removing publications written by firms), seven research

institutes and universities with ten published articles or more, are listed. Subsequently, a variable

consisting of the number of articles published in each county was computed. By doing so, we are able to

capture not only the existence of research institutions in proximity to the firms, but also the level of

productivity. This is important, as it is plausible to assume that research institutions with a higher

number of publications contribute more to innovation than an institution with fewer publications.

4.4.3 Independent variables: Higher education and general experience

The two hypotheses proposed in relation to the stock of human capital investigate the relationship

between level of formal- and informal education and its effect on impact of innovation.

Level of education is reported through the employee-employer dataset from SSB. By looking at the

highest completed level of higher education for all registered employees within the industry, detailed

8/13/2019 kinhte2

2/14

reports on individual educational level can be aggregated to firm-level variables. Other studies on this

topic have often been survey based and used ordinal scale variables (Ns et al. 1998; Dahl 2002;

Graversen et al. 2002) and grouped levels of education. We will, however, use the exact number of

years of higher education that the firm has accumulated divided by number of employees to calculate

the average level of higher education for each firm. The numbers of years are calculated based on

Norwegian Standard for grouping of Education Levels7.

As discussed previously, general experience covers informal and non-formal ways of acquiring

knowledge. Therefore, it is of value to capture the total knowledge base of employees that a firm has

accumulated through present and previous working experience. We chose not to limit this to knowledge

acquired through work within the offshore industry, but claim that knowledge gained through

experience in other industries also may be of value for the current employer and its ability to innovate.

As the formal education is captured by the measurement presented above, we calculate the general

experience by subtracting years used in formal education from each persons age. As the average years

of completed education in the offshore industry are 14 years and the vast majority of employees started

at school when they were 7 years

old8, 21 years is subtracted in the computation.

Further, we need to consider that experience may not follow a linear growth curve. By

applying a logarithmic function we acknowledge that there is a diminishing marginal

effect of gaining experience. Hence, the following equation is applied:

-21)

4.4.4 Independent variables: Intra- and inter industry mobility

These constructs aim at measuring the mobility of employees within the offshore

industry as well as mobility from other industries into the offshore industry. The

employer-employee database from Statistics Norway provides a detailed overview on

all employer-employee connections in Norway, as well as changes in these

connections. We use this database to compute firm-level mobility rates.

When looking at the relationship between employee mobility and impact of

innovation, we need to consider that it may exist a delay in the effect on the impact of

innovation due adjustments to a new working environment, training and other initial

activities. When computing these variables, we have introduced a two-year lag,

implying that the mobility-rates from 2006 have been used to measure Impact of

8/13/2019 kinhte2

3/14

innovation in 2008.

For intra-industry mobility, an employee who has moved within the industry the last 12

months is coded as an Intra-industry mover. For inter-industry mobility, an individual

that either has been unemployed, is a graduate, is coming from other industries, or

comes from the public sector the last 12 months is coded as an Inter-industry mover.

Based on this codification, the intra- and inter-mobility rates per firm are computed

from the number of new employees as a percentage share of the total workforce in

each firm.

4.4.5 Independent variable: Geographic proximity

A main problem of empirical studies on geographic proximity is to find a reliable operationalization of

geographic proximity. Various methods have been used in the research. A few recent studies have asked

respondents in surveys to identify important knowledge providers in close proximity as a measure

(Lublinski 2003; Ganesan et al. 2005). A problem with this approach is that firms themselves decide

which industry actors to include. Hence, a list can be insufficient and not necessarily provide a complete

overview over localization of industry actors. However, the majority of research conducted uses level of

agglomeration within certain geographic units, such as states or counties, as a measure of degree of

geographic proximity (e.g. Jaffe et al. 1993; Audretsch & Feldman 1996). A high level of agglomeration

indicates geographic proximity, which again is foundational for being considered a clustered region. An

advantage with data on e.g. county level is that it can provide complete and clear-cut measures of

relevant business activity within the county. A disadvantage with such a measure is that agglomerations

may span across county borders, and certain firms or locations can hence be omitted from the measure.

Based on the data available for this paper, geographic proximity will be measured by level of

agglomeration on county level. This will be operationalized using Balassa's (1965) index of revealed

comparative advantage. Both revenues and number of firms can be utilized as measures for this

purpose. However, we consider number of firms to be a better measure for geographic proximity as a

high number of firms in the same location indirectly imply close proximity.

Employing data from The Register of Business Enterprises, the average level of agglomeration for the

entire country will be estimated as number of business enterprises in the offshore industry over thenumber of enterprises in all industries. Next, using county (Norwegian: fylke) as the geographic unit, a

measure of agglomeration for each county will be estimated in relation to the level of agglomeration for

the whole country. In total, this provides an index where the level of agglomeration for each county,

comparable to the natural value of 1, will be the result.

Example for the level of agglomeration for the county Rogaland:

8/13/2019 kinhte2

4/14

# firm

#

#

#

of offshore s in Rogaland

of all firms in Rogaland

Level of agglomeration

of offshore firms in Norway

of all firms in Norway

The natural value of the index is 1, indicating that the level of agglomeration of the

county is equal to the average level of agglomeration of the country, hence no distinct

proximity. Thus, interpreting the Balassa-values, counties can obtain levels lower than

average (< Average1 standard deviation), at average (Average +/- 1 standard

deviation), and above average (>Average + 1 standard deviation), only the latter

indicating that the region is characterized by geographically proximate firms, and

hence can be classified as a clustered region.

4.4.6 Moderator variable: Clustering as a magnifier of stock of human capital and

mobility

As previously mentioned, there are few, if any, empirical works studying clustering as a

8/13/2019 kinhte2

5/14

magnifier of knowledge on innovation. Therefore, it is difficult to find good theoretical

support for our operationalization of these constructs. The relationship between the

dependent variable impact of Innovation and the independent variables are

hypothesized to be affected by level of agglomeration. As the level of agglomeration

changes the form of the relationship between the knowledge constructs (human

capital and employee mobility) and impact of innovation, moderating effects occur

(Hair et al. 2010).

This paper has suggested two constructs measuring stock of human capital, two

constructs measuring access to knowledge through mobility, and one independent

construct for geographic proximity. The latter is measured by level of agglomeration,

and used as a moderator variable for the four former. Founded on these, constructs

for the moderating effect of level of agglomeration are created as products of the

mean centered observations of each knowledge construct, and the mean centered

observations of level of agglomeration. Mean centered scores are used to avoid

problems with multicollinearity and make coefficients more interpretable (Jaccard &

Turrisi 2003, p.29). An example is

- - -

-

Intra industry mobility Intra industry mobility

Level of agglomeration Level of agglomeration

where the product is the moderating construct of intra-industry mobility and level of

8/13/2019 kinhte2

6/14

agglomeration. As mentioned earlier, high levels of agglomeration indicate clustering,

which moderated on the knowledge constructs, should result in magnifying effect on

impact of innovation.

4.4.7 Control variables

A control variable is a variable that is held constant and whose impact is removed in

order to analyze the relationship between other variables without interference. Firm

size, firm age, competition as well as R&D investment have been controlled for in this

study.

Firm size was measured by the number of employees in the firm. Prior studies have

identified a significant positive relationship between firm size and innovativeness. This

is also documented by the innovation survey performed by Statistics Norway.

Firm age was measured by the number of years since the founding of the firm. Prior

studies have identified a significant negative relationship between firm age and

innovativeness. Due to a new form of registration introduced in 1988, all firms

established before this date are coded with 1988 as their registration date in our

dataset. We have minimized this bias by applying a logarithmic transformation; hence

the relative importance of each year increased age is diminishing.

Competition follows a standard measurement used by Norwegian Competition

Authority. It is measured as the proportion of each firms revenue for its county,

squared and aggregated to county level, resulting in a scale ranging from 01. A low

number indicates high competition, and a high number indicates low competition.

R&D investment is measured by a survey question, asking respondents to report how much they invest

in R&D. The amount is normalied against the firms total turnover. Most previous studies have found a

positive relationship between R&D Investment and innovative performance.

A measure to control for tenure, i.e. the average number of years workers have been employed in the

firm was also created. However, introducing this control variable led to signs of multicollinearity in the

data set as this variable had high correlations with other measures, especially firm age and general

8/13/2019 kinhte2

7/14

experience. This implies that tenure to a large degree is overlapping with the other constructs

presented, and therefore not included in the final analysis.

In sum, the list of variables used in the regression analysis is as follows:

4.4.8 Data screening and transformations

Initial screening of the dataset shows that the dataset had characteristics of homoscedasticity, and no

problems with autocorrelation. However, the screening revealed that the residuals had a slight left

skewness. After analyzing the variables individually, we applied a square root-transformation on the

dependent variable to

normalize the residuals in the regression, in accordance with suggestions from the

literature (DeCoster 2001, p.10; Field 2009, p.220; Hair et al. 2010, p.80). In addition to

this, an analysis of the outliers in the dataset was performed, and non-normal

observations were deleted.

4.4 vn hnh ca cc bin

4.4.1 Bin phthuc: Tc ng ca si mi

Kho st i mi tthng k ca Na Uy sc sdng cung cp mt bin php ng tin cy ca s

i mi. Ph hp vi hng dn Oslo, SSB xc nh cc cng ty sng to nh "cng ty gii thiu mtdch vhoc sn phm mi hoc c ci thin ng khoc thc hin qu trnh ci thin ng k

trong giai on ny". Vic vn hnh i mi trong bo co ny c da trn mt cu hi trong cuc

kho st hi: "Bao nhiu phn trm doanh thu ca bn trong nm 2008 bt ngun tcc dch vhoc

sn phm ci thin ng k?" Ny cung cp mt bin php ng tin cy v chi tit vcc hot ng i

mi trong cc cng ty c kho st. Bng cch sdng bin php ny, chng ta c thnm bt khng

chhot ng i mi (v dnh mt sbng sng chslm g), nhng cng c thchp nh hng

ca i mi, nh chng ti cho i mi thc scung cp mt ngun thu nhp ca cng ty.

Trong nm 2008, cuc kho st i mi nhn c phn hi t614 cng ty trong ngnh cng nghip

ngoi khi. Sngi c hi trli cu hi lin quan n nhng g tlphn trm ca doanh thu xutpht tsng kin mi l 215.

4.4.2 bin c lp: ccn phm hc thut

Bin ny nhm mc ch o lng tc ng ca si mi bng nhng g trong cc cng ty c li t

vic nm gn vi cc tchc nghin cu. Sdng mt nh ngha rng, c nhng c sgio dc cung

cp gio dc c lin quan cho ngnh cng nghip ra nc ngoi trong 18 ca 19 qun trong Na Uy

8/13/2019 kinhte2

8/14

(Oljeindustriens Landsforening 2008). iu ny bao gm tt ccc nn gio dc c lin quan bng v cao

hn mc cnhn. Tuy nhin, sau Rothaermel & Ku (2008) l lun, c sgio dc (v dnh mt

trng Cao ng i hc) l khng cho vn tr tuc hng li i mi. Hcho rng nghin cu

cn phi c thc hin, v vy c mt skhc bit quan trng gia mt c sgio dc v tchc

nghin cu.

Cc tchc nghin cu c lin quan c la chn bi mt nh gi ca bi bo xut bn trong cc lnh

vc c lin quan trong tp ch chuyn ngnh. i vi mt bn tm tt chi tit vqu trnh ny, xem Ph

lc 2. Sau khi chnh sa bng tay (sp nhp cc tchc nghin cu bng v loi bccn phm c

vit bi cc cng ty), by vin nghin cu v cc trng i hc vi mi bi bo xut bn hay hn,

c lit k. Sau , mt bin bao gm slng cc bi bo xut bn trong mi qun c tnh

ton. Bng cch , chng ti c thnm bt c khng chstn ti ca cc tchc nghin cu

gn cc cng ty, nhng cng c mc ca nng sut. iu ny l quan trng, v n l hp l{ ginh

rng cc tchc nghin cu vi mt scao hn can phm ng gp nhiu hn cho si mi hn

mt tchc vi ccn phm t.

4.4.3 Cc bin c lp: Gio dc i hc v kinh nghim chung

Hai githuyt xut lin quan n cc cphiu vn con ngi iu tra mi quan hgia trnh hc

vn chnh thc v khng chnh thc v nh hng ca tc ng ca si mi.

Trnh hc vn c bo co thng qua bdliu nhn vin sdng lao ng tSSB. Bng cch nhn

vo mc hon thnh cao nht ca gio dc i hc cho tt ccc nhn vin ng k{ trong ngnh

cng nghip, cc bo co chi tit vtrnh hc vn c nhn c thc tng hp bin cng ty cp.

Cc nghin cu khc vchny thng c da kho st v sdng cc bin quy m tht(NAS

v cng snm 1998;. Dahl 2002;. Graversen v cng s2002) v cp nhm gio dc. Chng ti s,

tuy nhin, sdng cc con schnh xc ca nm ca gio dc i hc m cng ty tch ly c chiacho slao ng tnh ton mc trung bnh ca gio dc i hc cho mi cng ty. Cc con sca nm

c tnh ton da trn tiu chun ca Na Uy cho cc nhm Gio dc Levels7.

Nh tho lun trc , kinh nghim chung bao gm cch thc v khng chnh thc c c kin

thc. Do , n c gi trnm bt c tng sc stri thc ca nhn vin mt cng ty tch ly

qua kinh nghim lm vic hin ti v trc . Chng ti chn khng hn chny kin thc thu

c qua cng vic trong ngnh cng nghip nc ngoi, nhng cho rng kin thc thu c qua kinh

nghim trong ngnh cng nghip khc cng c thc gi trcho cc chnhn hin ti v khnng i

mi. Nh gio dc chnh thc bbt trong o lng trnh by trn, chng ta tnh ton kinh nghim

chung ca nm trsdng trong gio dc chnh thc tca mi ngi. Nh nhng nm trung bnh cagio dc hon thnh trong ngnh cng nghip nc ngoi l 14 nm v phn ln cc nhn vin bt u

trng khi h7 nm

old8, 21 nm c trtrong tnh ton.

Hn na, chng ta cn phi xem xt rng kinh nghim c thkhng theo mt ng cong tng trng

tuyn tnh. Qua

8/13/2019 kinhte2

9/14

p dng mt hm logarit chng ta tha nhn rng c mt bin gim dn

hiu qut c kinh nghim. Do , cc phng trnh sau y c p dng:

Kinh nghi -21)

4.4.4 Cc bin c lp: ni, lin ngnh cng nghip di ng

Cc cu trc nhm mc ch o lng sdi chuyn ca nhn vin trong ngoi khi

ngnh cng nghip cng nh tnh di ng tcc ngnh cng nghip khc trong ngnh cng nghip ngoi

khi. Cc

c sdliu sdng lao ng ca nhn vin tthng k ca Na Uy cung cp mt ci nhn tng quan chi

tit v

tt ccc kt ni sdng lao ng, ngi lao ng ti Na Uy, cng nh nhng thay i trong cc

kt ni. Chng ti sdng c sdliu ny tnh gi tnh di ng cng ty cp.

Khi nhn vo mi quan hgia tnh di ng ca nhn vin v tc ng ca

i mi, chng ta cn phi xem xt rng n c thtn ti mt schm trtrong cc hiu ng vtc

ng ca

i mi iu chnh do mi trng lm vic, o to mi v ban u khc

hot ng. Khi tnh ton cc bin ny, chng ti gii thiu mt khong hai nm,

ng rng tnh di ng, gi tnm 2006 c sdng o lng tc ng ca

i mi trong nm 2008.

Cho di ng trong ni bngnh cng nghip, mt nhn vin di chuyn trong cc ngnh cng nghip

cui cng 12

thng c m ho nh l mt 'trong ni bngnh cng nghip ng lc. Cho di chuyn qua nhiu

ngnh cng nghip, mt c nhn

m mt trong hai btht nghip, tt nghip, l n tcc ngnh cng nghip khc, hoc

xut pht tkhu vc nh nc trong 12 thng qua c m ha nh mt "ng lc lin ngnh cngnghip.

Da trn hthng ha ny, gi ni bv lin di ng ca mi doanh nghip c tnh

tslng nhn vin mi nh l mt tlphn trm ca tng slao ng trong

mi cng ty.

8/13/2019 kinhte2

10/14

4.4.5 c lp bin: gn gi a l

Mt vn chnh ca nghin cu thc nghim vmt a l{ l tm mt vn hnh ng tin cy ca s

gn gi va l{. Phng php khc nhau c sdng trong cc nghin cu. Mt vi nghin cu

gn y yu cu trli trong cuc iu tra xc nh cc nh cung cp kin thc quan trng trong

gn nh mt bin php (Lublinski 2003;. Ganesan v cng s2005). Mt vn vi phng php ny lcc cng ty tquyt nh m din vin ngnh cng nghip bao gm. Do , mt danh sch c thl

khng v khng nht thit phi cung cp mt ci nhn tng quan hon ton ni a ha ca cc din

vin ngnh cng nghip. Tuy nhin, phn ln cc nghin cu c tin hnh sdng cc mc tch t

trong cc n va l nht nh, chng hn nh cc quc gia hoc cc ht, nh mt thc o ca mc

gn gi va l (v dnh Jaffe v cng snm 1993;. Audretsch & Feldman 1996). Mt mc cao

ca stch tchra sgn gi va l, m li l nn tng c coi l mt khu vc tp trung. Mt li

thvi cc dliu trn v dcp huyn l n c thcung cp cc bin php y v r rng vhot

ng kinh doanh c lin quan trong qun. Mt bt li vi mt bin php nh vy l stch tc th

chiu di qua qun bin gii, v cc cng ty nht nh hoc cc a im do c thbqua cc bin

php.

Da trn cc dliu c sn cho bi bo ny, gn va l sc o bng mc tch ttrn cp qun.

iu ny sc vn hnh sdng (nm 1965) chsli thso snh ca Balassa. Cdoanh thu v s

lng doanh nghip c thc sdng nh l bin php cho mc ch ny. Tuy nhin, chng ta xem

xt slng ca cc cng ty l mt bin php tt hn v sgn gi va l{ nh mt slng ln ca

cc cng ty trong cng mt vtr gin tip gi ra gn.

Sdng dliu tSng k{ sn xut kinh doanh, mc trung bnh ca tch tcho ton bt nc s

c c tnh l slng doanh nghip kinh doanh trong ngnh cng nghip ngoi khi hn slng

doanh nghip trong tt ccc ngnh cng nghip. Tip theo, sdng qun (Na Uy: 'fylke') l n va

l, mt bin php tch tcho mi qun sc c tnh lin quan n mc tch ttrong cnc.

Trong tng s, iu ny cung cp mt chsny c mc tch tcho mi qun, so snh vi cc gi tr

tnhin ca 1, sc kt qu.

V dcho cc mc tch tca qun Rogaland:

# Cng ty

#

#

#

ca nc ngoi s trong Rogaland

ca tt ccc cng ty trong Rogaland

Mc tch t

8/13/2019 kinhte2

11/14

cc cng ty nc ngoi Na Uy

ca tt ccc cng ty trong Na Uy

Cc gi trtnhin ca chsl 1, chra rng mc tch tca

qun tng ng vi mc trung bnh ca tch tca t nc, do khng c khc bit

khong cch. V vy, gii thch Balassa-gi tr, qun c thc c thp hn mc

, mc trung bnh (trung bnh + / - 1 tiu chun - trung bnh (1 lch chun Trung bnh + 1 lch chun), chsau

chra rng khu vc ny c c im a l cc cng ty kcn, v

do c thc phn loi nh mt khu vc tp trung.

4.4.6 bin Moderator: Clustering nh mt knh lp cphiu ca ngun nhn lc v

di ng

Nh cp trc , c rt t, nu c, cng trnh thc nghim nghin cu phn nhm nh mt

knh lp kin thc vsi mi. V vy, rt kh tm thy tt l thuyt

htrcho vn hnh ca chng ta vcc cu trc. Mi quan hgia

tc ng bin phthuc ca i mi v cc bin c lp l

a ra githuyt bnh hng bi mc tch t. Nh mc tch t

thay i hnh thc ca mi quan hgia cc cu trc kin thc (con ngi

vn v tnh di ng ca nhn vin) v tc ng ca si mi, tc ng iu ha xy ra

(Tc et al. 2010).

8/13/2019 kinhte2

12/14

Bi vit ny nghhai cu trc o cphiu vn con ngi, hai

cu trc o tip cn tri thc thng qua di ng, v mt c lp

xy dng v sgn gi va l{. Sau ny c o bng mc tch t,

v sdng nh l mt bin iu tit cho bn cu. c thnh lp trn , xy dng

cho tc dng iu ha ca mc tch tc to ra nh cc sn phm ca

c ngha l quan st trung tm ca mi xy dng kin thc, v c ngha l tp trung

quan st mc tch t. C ngha l im trung tm c sdng trnh

vn vi a cng v lm cho hsthphin dch c nhiu hn (Jaccard &

Turrisi 2003, tr. 29) Mt v dl

---

-

Ngnh cng nghip di ng trong ni bngnh cng nghip di ng ni

Mc tch tMc tch t

ni m sn phm l xy dng iu ha ca tnh di ng trong ni bngnh cng nghip v mc

tch t. Nh cp trc , mc tch tcho nhm,

m kim duyt trn cc cu trc kin thc, nn kt qul phng i nh hng

tc ng ca si mi.

4.4.7 bin kim sot

Mt bin kim sot l mt bin c tchc lin tc v c tc ng c ly ra trong

t hng phn tch mi quan hgia cc bin skhc m khng cn can thip. Cng ty

kch thc, tui cng ty, cnh tranh cng nh u t R & D c kim sot trong ny

8/13/2019 kinhte2

13/14

nghin cu.

Quy m doanh nghip c o bng slng nhn vin trong cng ty. Cc nghin cu trc khi c

xc nh mt mi quan htch cc gia quy m doanh nghip v tnh sng to. Ny

cn ti liu ca cuc kho st c thc hin bi si mi Thng k Na Uy.

Tui cng ty c o bng snm ktngy thnh lp ca cng ty. Trc

nghin cu xc nh c mt mi quan htiu cc gia tui cng ty v

sng to. Do mt hnh thc mi ng k{ c gii thiu trong nm 1988, tt ccc cng ty

thnh lp trc ngy ny c m ho vi nm 1988 nh l ngy ng k{ ca htrong ca chng ti

bdliu. Chng ti gim thiu thin vny bng cch p dng mt bin i logarit, v vy

tm quan trng tng i ca mi nm tui tng ang gim dn.

Cnh tranh theo mt tiu chun o lng c sdng bi cnh tranh Na Uy

Thm quyn. N c o bng tldoanh thu mi ca cng ty cho cc qun ca n,

bnh phng v tng hp cho cp qun, dn n mt quy m khc nhau, 0-1. Mt thp

chscnh tranh cao, v mt slng ln cho thy cnh tranh thp.

u t R & D c o bng mt cu hi kho st, yu cu ngi trli cho bit hu t bao nhiu vo

R & D. Stin l bnh thng so vi tng doanh thu ca cng ty. Nghin cu trc y hu ht tm

thy mt mi quan htch cc gia u t R & D v hot ng sng to.

Mt bin php kim sot cho nhim kz, tc l slng trung bnh cc nm cng nhn c s

dng trong cc cng ty cng c to ra. Tuy nhin, gii thiu cc bin ny kim sot dn n du

hiu ca a cng trong cc dliu thit lp nh l bin ny c mi tng quan cao vi cc bin php

khc, c bit l tui cng ty v kinh nghim chung. iu ny ng rng nhim kzn mt mc ln

l chng cho vi cc cu trc khc c trnh by, v do khng bao gm trong phn tch cui cng.

Tm li, danh sch cc bin sdng trong phn tch hi quy nh sau:

4.4.8 sng lc v chuyn i dliu

Kim tra ban u ca cc bdliu cho thy rng cc dliu c c im ca homoscedasticity, v

khng c vn vi ttng quan. Tuy nhin, vic kim tra cho thy cc sd c skewness tri nh.

Sau khi phn tch cc bin c nhn, chng ti p dng mt gc-chuyn i vung trn bin phthuc

bnh thng ha cc sd trong hi quy, ph hp vi cc xut ca

vn hc (DeCoster nm 2001, trang 10; trng nm 2009, p.220;. tc v cng snm 2010, p.80). Ngoi

8/13/2019 kinhte2

14/14

ny, mt phn tch vgi trngoi lai trong cc bdliu c thc hin, v khng bnh thng

quan st bxa....