i
Unde s anding S uden s’ Academic Achie emen
in Public High School
Ana Filipa Rosa Lou o
E idence o Po ugal
Disse a ion as pa ial equi emen o ob aining he
Mas e ’s deg ee in In o ma ion Managemen
i
MEGI
2018
Unde s anding S uden s’ Academic Achie emen in Public High School
E idence o Po ugal
Ana Filipa Rosa Lou o
MGI
i
ii
NOVA In o ma ion Managemen School
Ins i u o Supe io de Es a ís ica e Ges ão de In o mação
Uni e sidade No a de Lisboa
UNDERSTANDING STUDENTS’ ACADEMIC ACHIEVEMENT IN PUBLIC
HIGH SCHOOL: EVIDENCE FOR PORTUGAL
by
Ana Filipa Rosa Lou o
Disse a ion p esen ed as pa ial equi emen o ob aining he Mas e ’s deg ee in In o ma ion
Managemen , wi h a specializa ion in Knowledge Managemen and Business In elligence
Ad iso : F ede ico Miguel Campos C uz Ribei o de Jesus
Co Ad iso : Jo ge Nelson Gou eia de Sousa Ne es
May 2018
iii
ACKNOWLEDGEMENTS
This hesis would no be possible wi hou he help o all ha a e dea o me and also hose ha
wo ked beside me.
To my pa en s, my b o he and my boy iend, I’m especially g a e ul o all he pa ience and suppo
du ing he accomplishmen o his hesis.
To my iends ha always encou aged me o do my bes .
To my ad iso , P o esso F ede ico C uz Jesus, o all he help, suppo and wise wo ds. Thank you o
you ad ices, con ibu ions and cons uc i e eedback.
To P o esso Jo ge Ne es o all he ad ices and a ailabili y.
To Di eção-Ge al de Es a ís icas da Educação e Ciência, in pa icula , o D a. Joana Dua e and D a.
Ca a ina A lalo o all he a ailabili y and specially o c ea ing he condi ions o de elop his wo k
and p o ide me he access o he da a, c ucial o my wo k.
I was a pleasu e o walk his pa h wi h all o you.

i
ABSTRACT
Se e al pape s and s udies ha e been conduc ed o be e unde s and wha a e he main ac o s ha
in luence s uden s’ academic achie emen and wha measu es should be aken o imp o e i .
The e o e, based on 383.560 s uden s’ obse a ions, e alua ed on seconda y Po uguese public
schools in 2014/2015 academic yea , he pu pose o his s udy is o p o ide a new app oach o he
collec ed da a by using Da a Mining p edic i e models. The esul s show di e ences on he academic
achie emen among emales and male s uden s, whe e emales go be e academic esul s. Access
o compu e and In e ne ound o be powe ul ools in educa ion ha s uden s can explo e o hei
bene i and show o ha e a posi i e in luence on academic esul s. S uden s bene i ing om inancial
social suppo p o e o ha e a lowe pe o mance in academic achie emen . Resul s also poin o he
ac ha he numbe o ep o es s ill has a g ea nega i e impac on s uden s’ academic
achie emen . This is one o he i s s udies o he bes o he au ho s knowledge o employ analy ic
echniques on such a la ge da ase on he con ex o academic achie emen .
KEYWORDS
Academic Achie emen ; P edic i e Models; Educa ion
INDEX
1. In oduc ion .................................................................................................................. 1
2. Theo e ical backg ound ................................................................................................ 2
2.1. The concep o academic achie emen ................................................................. 2
2.2. P io esea ch on academic achie emen ............................................................. 2
3. Concep ual Model o unde s anding academic achie emen .................................... 9
4. Me hodology and esul s ........................................................................................... 12
4.1. Da a ..................................................................................................................... 12
4.2. Desc ip i e S a is ics and non-pa ame ic es s .................................................. 12
4.3. Decision T ees...................................................................................................... 18
5. Discussion ................................................................................................................... 26
5.1. Discussion o Findings.......................................................................................... 26
5.2. P ac ical implica ions ........................................................................................... 27
5.3. Theo e ical implica ions ...................................................................................... 27
6. Conclusions ................................................................................................................. 28
7. Limi a ions and ecommenda ions o u u e wo ks ................................................. 29
8. Bibliog aphy ................................................................................................................ 30
9. Appendix 1 .................................................................................................................. 36
i
LIST OF FIGURES
Figu e 4.1 – Final Classi ica ion do no ollow a no mal dis ibu ion ...................................... 13
Figu e 4.2 – Decision T ee o academic achie emen a cou se le el (Model 1) ................... 20
Figu e 4.3 – Decision T ee o academic achie emen a yea le el (Model 2) ....................... 23
Figu e 4.4 – Cumula i e Li and Cumula i e Cap u ed Response o Model 1 ....................... 25
Figu e 4.5 – Cumula i e Li and Cumula i e Cap u ed Response o Model 2 ....................... 25
Figu e 9.1 – SAS MINER MODEL 1 ............................................................................................ 36
Figu e 9.2 – SAS MINER MODEL 2 ............................................................................................ 37
ii
LIST OF TABLES
Table 2.1 – Re iew o p io esea ch on academic achie emen .............................................. 8
Table 4.1 – No mali y Tes ....................................................................................................... 13
Table 4.2 – S uden s’ Cha ac e is ics ....................................................................................... 15
Table 4.3 – Pa en s’ Socioeconomic Cha ac e is ics ................................................................ 16
Table 4.4 – Schools’ Cha ac e is ics ......................................................................................... 17
Table 4.5 – Cou ses by Gende ................................................................................................. 18
6
Teache s’ Cha ac e is ics
The li e a u e also epo s o be ai ly easonable, h ough eg ession analysis, o assume ha
eache s’ in luence is among he mos signi ican de e minan o explain he s uden s’ academic
achie emen (Rocko , 2004) leading o an eme gen in e es and he e o e o a g ow h on he
numbe o s udies on how eache s’ cha ac e is ics a ec he s uden s’ academic achie emen
(Buddin & Zama o, 2009; Clo el e , Ladd, & Vigdo , 2006; Goldhabe & Hansen, 2013; Gua ino,
Reckase, S acy, & Woold idge, 2015; Ri kin e al., 2005). I is wi hou su p ise ha Hanushek (2011)
epo s eache s as one he mos c ucial ac o s o s uden s’ academic achie emen . Howe e , i is
impo an o know which cha ac e is ics mos likely explain he eache ’ impac on s uden s’
academic achie emen . A ecen esea ch conduc ed in Po ugal (Sousa, Po ela, & Sá, 2003), using
da a om he pe iod be ween 2010 and 2012, s udied he impac o gende , eache si ua ion,
educa ion le el and expe ience by using as me hodology OLS eg ession analysis, concluded ha
emale eache s ha e highe in luence on s uden s’ academic achie emen han males’ eache s and
ha eache s wo king away om home ha e signi ican nega i e e ec s on s uden s’ academic
achie emen . Ad anced deg ee eache s (Mas e s o PhDs) seemed o ha e no e ec on he lowe o
highe s uden s’ pe o mance when compa ed o hose wi h a g adua ion deg ee. Finally, i ’s also
poin ed ha eache s wi h mo e expe ience a e mo e e ec i e inc easing s uden achie emen gains
han hose wi h less expe ience. The eache cha ac e is ics p esen ed p e iously sugges ed ha
he e’s a posi i e co ela ion be ween eache expe ience on ma h and eading esul s when applied
mos ly OLS eg ession analysis (Clo el e e al., 2006; C oninge , Rice, Ra hbun, & Nishio, 2007). In
addi ion, s uden s augh by emale eache s sco ed signi ican ly highe han hose augh by male
eache s in bo h ma hema ics and science as conclude Wößman (2003) a e applying WLS
eg essions. The posi ion o eache educa ion le el is an aspec ha does no p esen a clea
consensus, since he e a e di e en conclusions. Fo ins ance, C oninge (2007) epo s posi i e
e ec s be ween eache s’ educa ion le el and s uden s’ achie emen , howe e , Ri kin (2005) aise
se ious doub s on his opic.
As displayed in he able below, he e a e ew s udies applying s uden s, pa en s, schools and
eache ’s cha ac e is ics o explain s uden s’ academic achie emen . As a way o su pass his lack o
in o ma ion, he pu pose o his s udy is o analyze each o hese ou dimensions a he same ime,
al hough he e is no in o ma ion abou eache s’ cha ac e is ics as men ioned be o e.

7
Re
Da a
Me hods
S uden s
Pa en s
Schools
Teache s
Findings
(Hanushek &
Kimko, 2000)
Cogni i e skills o 39 coun ies, only 31 coun ies
ha e he measu emen o economic pe o mance
Reg ession models
x
x
• In e na ional ma hema ics and science es sco es a e s ongly ela ed o g ow h o na ions.
• Di ec spending on schools has no ela ionship o s uden pe o mance di e ences.
• Home-coun y quali y di e ences o immig an s a e di ec ly ela ed o U.S. ea nings.
• Ma hema ics and science skills a e ele an o he labo o ce.
(Hoxby, 2000)
Connec icu , USA: 649 elemen a y schools wi h da a
om 1992-1993 o 1997-1998 and 146 elemen a y
dis ic s wi h da a om 1986-1987 o 1997-1998
Reg ession models
x
x
• Class size does no ha e a s a is ically signi ican e ec on s uden achie emen .
• Class size educ ion has g ea e e ec in schools wi h mo e low income o A ican-Ame ican s uden s.
• Policy expe imen s con aining incen i es p oduce be e esul s han class educ ion.
(Fan & Chen,
2001)
Me a-analysis om 25 di e en s udies
Gene al linea model
(GLM)
x
x
• Posi i e ela ionship be ween pa en al in ol emen and s uden s’ academic achie emen , when applying GPA.
• Pa en al home supe ision has e y low ela ionship wi h s uden s’ academic achie emen .
• S ong ela ionship be ween pa en s’ aspi a ion/expec a ion and s uden s’ academic achie emen
• Low ela ionship be ween pa en al home supe ision and s uden s’ academic achie emen .
(Ba ne e al.,
2002)
152 seconda y schools om No he n I eland,
be ween 1994-1995 and 1995-1996 academic yea s
Linea P og amming
echniques
x
• Posi i e ela ionship be ween e ec i eness-e iciency pe o mance sco es and seconda y school size.
• La ge seconda y schools pe o m be e han smalle ones.
(Rocko ,
2004)
10,000 elemen a y-school s uden s and 300
eache s om wo dis ic s in New Je sey. In dis ic
A be ween 1989-1990 o 2000-2001 and dis ic B
be ween 1989-1990 o 1999-2000 academic yea s
Reg ession models
x
• La ge di e ences in quali y among eache s wi hin schools.
• Teache s’ expe ience inc eases s uden es sco es, pa icula ly in eading subjec a eas.
(D iessen e
al., 2005)
P ima y school om he Ne he lands, wi h mo e
han 500 schools and 12,000 s uden s, in 1994-1995
academic yea
F equency, Va iance and
S uc u al models
x
x
x
• No di ec e ec o pa en al in ol emen on s uden s’ academic achie emen .
• No di ec e ec on schools wi h nume ous mino i y pupils whe e hey appea o p o ide a conside able amoun o ex a e o wi h
espec o pa en al in ol emen .
(Ri kin e al.,
2005)
Public school s uden s om Texas. Da a o h ee
coho s be ween 1993-1995 academic yea
Reg essions models
x
x
• Class size educ ion is no a good p edic o o explain s uden s’ academic achie emen .
• Teache is an impo an ac o o explain school quali y.
(A chibald,
2006)
Elemen a y schools om Ne ada, USA, wi h mo e
han 60,000 s uden s, be ween 2002-2003 academic
yea
Hie a chical linea
models (HLM)
x
x
x
• Teache pe o mance is posi i ely ela ed o s uden s’ academic achie emen .
• Pe -pupil expendi u e a he school le el is posi i ely ela ed o s uden s’ academic achie emen in eading as i indica es wha
esou ces ma e o educa ion.
• S uden backg ound cha ac e is ics ma e , a he s uden le el and school le el.
• School size and school le el po e y ha e nega i e impac s on bo h ma h and eading esul s.
(Jackson e al.,
2006)
140 child en om USA, be ween Decembe 2000
and June 2002 wi h an a e age age o 13.8 yea s
In e ne eco ded
x
• Child en using mo e in e ne ha e be e esul s in eading achie emen han child en who used i less.
• Despi e he age, he use o in e ne has no e ec on s uden s’ academic pe o mance.
(J.-S. Lee &
Bowen, 2006)
415 child en o 3 d un il 5 h g ade om he
sou heas e n Uni ed S a es in 2004 academic yea
Hie a chical linea
models (HLM)
x
x
• Pa en s wi h di e en demog aphic cha ac e is ics and di e en ypes o in ol emen om dominan g oups had he s onges
associa ion wi h achie emen .
• Pa en al homewo k help was nega i ely associa ed wi h Eu opean Ame ican s uden s’ academic achie emen .
• Pa en in ol emen a school and high educa ional expec a ions, displayed he s onges ela ionship wi h achie emen .
(Ma ks e al.,
2006)
PISA 2000, o e 6,000 schools ac oss 32 coun ies
I em Response Theo y
(IRT)
Reg ession models
x
x
x
• Cul u al ac o s show o be impo an o explain socioeconomic inequali ies in educa ion.
• Cul u al esou ces play a mo e impo an ole in socioeconomic inequali ies in s uden s’ academic achie emen han ma e ial esou ces
a home.
• Ma e ial and lea ning in as uc u e play a mo e impo an ole o s uden pe o mance in ma hema ics and science han o eading.
(Jeynes, 2007)
Me a-analysis, om 52 s udies, be ween 1972-2000
Reg ession models
x
• Pa en al in ol emen has a posi i e impac on seconda y s uden s’ academic achie emen .
(Codjoe, 2007)
Sample om black s uden s in Edmon on, Canada
In e iews
x
• Home en i onmen and pa en al suppo con ibu e o s uden s’ academic achie emen .
(C oninge e
al., 2007)
F om Ea ly Childhood Longi udinal S udy,
Kinde ga en Class o 1998–1999
Hie a chical linea
models (HLM)
x
x
• Teache s’ deg ee ype and expe ience posi i ely a ec s uden s’ eading achie emen .
• Teache s’ quali ica ions in luence s uden s’ academic achie emen s on eading and ma hema ics.
(H. Lee, 2007)
80 high schools and 52 middle schools, om USA,
wi h s uden s g ades 7 o 12, in 1994
Hie a chical linea
models (HLM)
Classic lineal eg ession
model
x
x
x
• Pee acial/e hnic composi ion do no media e he ela ionship be ween school acial/e hnic composi ion and achie emen .
• Racial/e hnic composi ion o schools ma e s o educa ional achie emen in he USA.
(Lei & Zhao,
2007)
Middle school om Ohio, USA wi h 237 s uden s,
be ween 2003–2004 academic yea
Hie a chical linea
models (HLM)
ANOVA es s
x
• The quan i y o echnology use, pe i sel , is no c i ical o s uden lea ning.
• When he quali y o echnology use is no ensu ed, mo e ime on compu e s may cause mo e ha m han bene i .
• When GPA changes, echnology wi h highe impac on s uden s we e hose ela ed o speci ic subjec a eas and s uden de elopmen .
8
(S einmay &
Spina h, 2008)
342 s uden s om a Ge man school in 11 h and 12 h
g ade s
Reg essions models
x
• Gende di e ences a e p esen ed in mos o he a iables s udied.
• Gi ls’ g ades we e significan ly be e han boys’.
• Pe sonali y and mo i a ion play impo an oles in gende di e ences in school a ainmen .
• School a ainmen is a be e p edic o o gi ls han o boys o explain gende di e ences in academic achie emen .
(Ca o e al.,
2009)
Canada’s Na ional Longi udinal S udy wi h a sample
o 6290 s uden s be ween 1994-2001 academic
yea s
Hie a chical linea
models (HLM)
Panel da a models
x
• Highe disc epancy in ma hema ics achie emen among s uden s wi h highe and lowe SES amilies.
(Mensah &
Kie nan, 2010)
Millennium Coho S udy, wi h child en in he
p ima y yea o school, England, be ween 2005-
2006 academic yea
Tobi eg ession models
Uni a ia e and
Mul i a ia e analyses
x
x
• S uden s socioeconomic disad an ages show lowe a ainmen in communica ion, language and li e acy, and ma hema ical
de elopmen .
• Ea ly mo he hood, low ma e nal quali ica ions, low amily income and unemploymen p edic lowe sco es a school.
• Gende di e ences a e iden i ied o s uden s in amilies whe e: mo he s a e young, lack o ma e nal quali ica ions, o hey a e li ing in
poo quali y a eas.
(Hanushek,
2011)
Hanushek and Ri kin (2010)
Reg ession models
x
• Posi i e co ela ion be ween eache s’ e ec i eness and ma ginal gains in s uden s’ u u e ea nings.
(Ha as, 2011)
Longi udinal sample om Millennium Coho S udy,
om England, o child wi h 3 and 5 yea s
Uni a ia e analyses o
a iance
Chi-squa e es s
x
• Social-economic s a us does no a ec pa en s’ pa icipa ion in lea ning ac i i ies.
• Families’ income and pa en s’ educa ion ha e a s ong e ec on child en's language/li e acy (ma e nal educa ion has a s onge e ec ).
• Socioeconomic disad an age and lack o ma e nal educa ional quali ica ion s ongly in luence child en compe encies.
(Pa e son &
Pahlke, 2011)
Public middle school, in he sou hwes e n Uni ed
S a es, wi h 211 s uden s, be ween 2007-2011
academic yea s
Reg ession models
x
x
• S uden cha ac e is ics a e associa ed wi h s uden s’ academic achie emen .
• A ican Ame ican and La ina s uden s end o ha e lowe g ades han o he s uden s.
• P io achie emen show o be a signi ican p edic o o s uden s’ academic achie emen .
• Gende s e eo yping is a signi ican p edic o o s uden s’ academic achie emen .
(Hanushek &
Woessmann,
2012)
64 di e en coun ies be ween 1964 and 2003 yea s
Reg ession models
x
x
• School policy can be a key ins umen o spu g ow h.
• Di e ences in cogni i e skills lead o di e ences in economic g ow h.
(B unne e al.,
2013)
PISA 2003, wi h 275,369 15 h yea s old s uden s
om 41 na ions
Mul iple g oup ac o
analy ic models
Full maximum likelihood
me hod “MLR”
x
• Gi ls ou pe o med boys in eading achie emen in all coun ies s udied.
• Boys ou pe o med gi ls in ma hema ics achie emen in almos all coun ies s udied.
• A ully hie a chical concep ualiza ion o achie emen , con ibu es o a be e unde s anding o gende di e ences.
(Wally-Dima &
Mbekomize,
2013)
660 S uden s om Bachelo o Accoun ancy deg ee
p og am a he Uni e si y o Bo swana in 2011-2012
academic yea
Desc ip i e s a is ics T
es s
x
• Indi idual’s commi men and igh a i ude owa d accoun ing s udies a e he key ac o s o explain academic pe o mance.
• Female s uden s pe o m be e han male s uden s.
(Boswo h,
2014)
Public school om No h Ca olina, USA, wi h 4 h
and 5 h g ade s uden s, o 2000-2001 academic
yea
Reg ession models
x
x
• S uden s a e assigned o class ooms based on s uden s’ cha ac e is ics (Gende , E hnici y, Pa en s educa ion, o he s).
• S uden s who s uggle in school bene i mo e om class size educ ions when compa ed wi h hose on he op o he achie emen
dis ibu ion.
• Smalle classes ha e smalle achie emen gaps.
• Class size educ ion is mo e e ec i e a closing achie emen gaps han aising achie emen .
• Class size e ec s on bo h a e age achie emen and achie emen gaps a e small.
(K assel &
Heinesen,
2014)
Seconda y school om Denma k wi h s uden s o
9 h and 10 h g ade, be ween 2003-2006 academic
yea s
Reg ession discon inui y
design (RDD)
Con ol o school ixed
e ec s (SFE)
O dina y Leas Squa es
(OLS)
x
x
x
• Nega i e e ec s o class size on s uden s’ academic achie emen .
(Vigdo e al.,
2014)
Public school s uden s om 5 h o 8 h g ade, in
No h Ca olina, be ween 2002-2005 academic yea s
P obi eg ession
Reg ession models
x
• Home compu e echnology is associa ed wi h nega i e impac s on s uden ma h and eading sco es.
• P o iding uni e sal access o home compu e s and high-speed in e ne access would b oaden, a he han na ow, ma h and eading
achie emen gaps.
(Hodis e al.,
2015)
Seconda y schools, om New Zealand, wi h a
sample o 782 s uden s
Hie a chical linea
models (HLM)
x
• Maximal le els o aspi a ion and minimal bounda y goals p edic s uden s’ academic achie emen .
• Maximal le els o aspi a ion, minimal bounda y goals and s uden s’ academic achie emen a e mode a ed by he ype o assessmen
asks.
(C. L. Lee &
Mallik, 2015)
S uden s om he Uni e si y o Wes e n Sydney,
be ween 2007–2012 academic yea s
O dina y Leas Squa es
(OLS)
x
• Posi i e associa ion be ween uni e si y en y sco es and s uden s’ academic achie emen .
• S uden pe o mance is ela ed o age and s uden s’ g ades.
Table 2.1 – Re iew o p io esea ch on academic achie emen
9
3. CONCEPTUAL MODEL FOR UNDERSTANDING ACADEMIC ACHIEVEMENT
The li e a u e e iew conduc ed on he p e ious sec ion allowed us o be awa e o he main
an eceden s o academic achie emen . By combining he esul s o mul iple pas s udies and heo ies
ha suppo ed hem, we ha e buil a comp ehensi e esea ch model o shed some ligh on wha
d i es academic achie emen . Based on he li e a u e, we ha e iden i ied ou con ex s ha may
a ec academic achie emen , namely he cha ac e is ics o s uden s, pa en s, schools and eache s.
Howe e , as he e a e con adic o y indings on he li e a u e, and due o da a a ailabili y,
men ioned below, he las one ( eache s) was excluded om he con ex o his s udy. Hence, wi hin
each o he h ee cons uc s ha a e likely o in luence academic achie emen , some ela ionships
a e hypo hesized.
Gende di e ences is one o mos s udied cha ac e is ics o e he yea s. In ac , he main conclusions
e e ha emale s uden s ob ain be e academic esul s when compa ed wi h male s uden s
(Mensah & Kie nan, 2010; S einmay & Spina h, 2008; Wally-Dima & Mbekomize, 2013) despi e
hese conclusions being mo e p onounced in some academic a eas (B unne e al., 2013; Ghaz ini &
Khajehpou , 2011; S einmay & Spina h, 2008). Usually emales end o ha e be e academic
pe o mance. The e o e, we hypo hesize:
H1: Gende will ha e an impac on s uden s’ academic achie emen as emales will pe o m be e .
S uden s’ sociodemog aphic cha ac e is ics, mo e speci ically i he s uden is na i e (has he same
na ionali y as he coun y unde s udy) o immig an (has o he na ionali y), ha e been s udied om
he momen na i e s uden s ha e p esen ed be e esul s han immig an ones (S and, 2011). In
his s udy, will only be p esen ed i he s uden has a Po uguese na ionali y o o he . The e o e, we
hypo hesize:
H2: Na i e s uden s will pe o m be e on academic achie emen .
Compu e access is ano he cha ac e is ic ha igge s di e en conclusions. In his con ex some
esea che s a e less op imis ic abou he ela ionship be ween academic achie emen and access o
compu e s (Lei & Zhao, 2007; Vigdo e al., 2014; Wen wo h & Middle on, 2014), while he mo e
op imis ic conside ha compu e s as a wo king ool a e a bene i o s uden s since we cu en ly li e
in a digi al age and in o ma ion sys ems (Bo zekowski & Robinson, 2012; Gil-Flo es, 2009; Lei & Zhao,
2007; Vigdo e al., 2014). The e o e, we hypo hesize:
H3: S uden s wi h compu e access will pe o m be e on academic achie emen .
In e ne access is one o he cha ac e is ics ha has gene a ed mo e con adic ions o less clea
conclusions since i can be seen as a dis ac ion, when used excessi ely (Kubey e al., 2001; Liebe &
Chou, 2001), and no o academic pu poses, bu on he o he hand, i can be seen as an added alue
o he s uden s, p o iding a wide lea ning ne wo k (Jackson e al., 2006; To es-Díaz e al., 2016).
The e o e, we hypo hesize:
H4: S uden s wi h in e ne access will pe o m be e on academic achie emen .
10
Al hough no e e ence was ound be ween he numbe o p e ious s uden s ep o e yea s and he
academic achie emen in his s udy, wha is being s udied is whe he he ac ha he s uden ha ing
ep o ed in p e ious yea s o no could in ac be a ac o ha in luences s uden s’ academic
achie emen . The e o e, we hypo hesize:
H5: S uden s ha ha e ep o ed in he pass will p esen lowe le els on academic achie emen in
he u u e.
Family suppo om Social Se ices (SASE) is a social bene i which he main pu pose is suppo ing
unde p i ileged amilies who ha e child en o school age, gua an eeing equal access oppo uni ies
and school success o all s uden s in p ima y and seconda y educa ion le els. I also ies o p omo e
socio-educa ional suppo measu es o he s uden s o households whose economic si ua ion
de e mines he need o inancial con ibu ions o school expenses such as he pu chase o books
and school supplies, meals and anspo (DGE, 2018). Wha we wan o ind ou is whe he he
s uden s ha ecei e his kind o suppo , will be impac ed on hei academic achie emen .
The e o e, we hypo hesize:
H6: S uden s ha ecei e suppo om social se ices (SASE) will ha e lowe le els on academic
achie emen .
Family inancial suppo is a social bene i a ibu ed mon hly o amilies. The objec i e is o
compensa e households’ expenses ela ed o he sus enance and educa ion o child en and young
people. Wha we wan o ind ou is whe he he s uden s ha ecei e his kind o suppo , will be
impac ed on hei academic achie emen (Segu ança Social, 2018). The e o e, we hypo hesize:
H7: S uden s ha ecei e amily inancial suppo will ha e lowe le els on academic achie emen .
Mo he educa ion le el e e s o he le el o he academic deg ee o he mo he and is one o he
cha ac e is ics ha p esen a clea impac on academic achie emen (Ha as, 2011; Mensah &
Kie nan, 2010), being also one o he a iables ha is di ec ly linked o he pa en al SES (Ca o e al.,
2009; Si in, 2005; S einmay e al., 2010). Fa he educa ion le el e e s o he le el o he academic
deg ee o he a he and is one o he cha ac e is ics ha seems clea o be in e es ing o know how
can impac he academic achie emen , being also one o he a iables ha is di ec ly linked o he
pa en al SES (Ca o e al., 2009; Si in, 2005; S einmay e al., 2010). The e o e, we hypo hesize:
H8: Pa en s educa ion le el will ha e a posi i e impac on academic achie emen .
Class size is he numbe o s uden s pe class. The e a e se e al (and o some ex en con adic o y)
conclusions on his opic di e ging when i comes o an o e all ag eemen . Al hough he e a e
esea che s who de end ha he e is no di ec ela ionship be ween he educ ion o numbe o
s uden s pe class and he inc ease o s uden s’ academic achie emen (Hoxby, 2000; Wößmann &
Wes , 2006), he e a e hose who a gue ha s uden s bene i a lo when implemen ing his measu e
(Boswo h, 2014; K assel & Heinesen, 2014; Ri kin e al., 2005). The e o e, we hypo hesize:
H9: Class size will ha e a nega i e impac on academic achie emen .
11
School size is he cha ac e is ic ha measu es he numbe o s uden s pe school. E en hough he e
migh be no di ec connec ion be ween he s uden s’ academic achie emen and school size, he e
a e hose who de end ha la ge schools ha e be e s uden ´s esul s (Ba ne e al., 2002). The
u h is ha a link be ween he dimension o he schools and s uden s’ esul s can be ound: bigge
schools will ha e bigge classes. Wi h his in mind, we ound e idences ha p o e he opposi e: a
nega i e ela ionship be ween school size and s uden s’ academic achie emen (A chibald, 2006;
Egali e & Kisida, 2016; Welsch & Zimme , 2016). The e o e, we hypo hesize:
H10: School size will ha e a posi i e impac on academic achie emen .

12
4. METHODOLOGY AND RESULTS
4.1. DATA
To each he p oposed objec i es in his s udy, wi h he mos eliable and comple e da a, we used
da a om MISI da a base. The MISI da abase is he in o ma ion sys em whe e educa ional da a
conce ning p e-schola , basic, seconda y om public schools unde MEC (Minis y o Educa ion and
Science) and some ypes o p i a e schools is collec ed. I s pu pose is o cen alize all educa ional
da a collec ion om p e-schools, basic and seconda y, as well as p o ide o he espec i e ins i u es
he necessa y in o ma ion ha will se e as basis o he p oduc ion o educa ional s a is ics o he
decisions-making p ocesses. The public educa ion con ex comp ises ou p og ams: employees,
accoun ing, s uden s and school social ac ions. In he con ex o his pape , he MISI da a base was
used o collec all he da a om s uden s, pa en s and schools, howe e we we e no able o e ie e
da a om eache s esul ing in a limi a ion o his s udy. To be e con ex ualize he da a collec ed,
we also used da a om Po uguese Na ional Ins i u e o S a is ics (INE) o ga he in o ma ion on he
s uden s’ esidence a ea, speci ically on popula ion densi y, mon hly a e age income, pe cen age
a e age on cul u e expenses, aging index, esidence popula ion and unemploymen a e.
All da a om he MISI da abase ega ding s uden s, pa en s and schools was collec ed a he DGEEC
acili ies in Lisbon be ween No embe 2016 and Janua y 2017. P og amming echniques we e used
o collec hem, in his case SQL, namely SQL Se e Managemen S udio ool. Fo he ea men o
he da a collec ed, da a analysis echniques we e used, ecu ing o SAS so wa e, mo e speci ically
SAS Guide and SAS Mine ools.
A e he app op ia e da a p ocessing i , was also added da a om INE sou ce o be e con ex ualize
he da a. The inal da abase con ains a o al o 383560 obse a ions, om Po uguese s uden s
a ending public schools in he yea s 2014/2015 o 10 h, 11 h and 12 h g ades e alua ed and
a ending he 21 cou ses whe e na ional exams ake place, since was excluded Po uguese cou se
ese ed o s uden s wi h se e ed o p o ound dea ness, so ha he analysis would be he mos
accu a e as possible as hese comp ise he majo i y o egis e ed s uden s (Gabine e do Sec e á io
de Es ado da Educação, 2017), his means: Biology and Geology, D awing A, Economics A,
Philosophy, Physics and Chemis y A, Geog aphy A, Desc ip i e Geome y A, His o y A, His o y B,
His o y o Cul u e and A s, La in A, Po uguese Li e a u e, Fo eign Language - Ge man, Fo eign
Language - Spanish, Fo eign Language - F ench, Fo eign Language - English, Ma hema ics A,
Ma hema ics Applied o Social Sciences, Ma hema ics B, Po uguese and Po uguese Non Ma e nal
Language.
4.2. DESCRIPTIVE STATISTICS AND NON-PARAMETRIC TESTS
To be e unde s and he conclusions ound on he li e a u e e iew i is i al o pe o m es s so ha
ou analysis can be as much accu a e and comple e as possible. Howe e , i is impo an o
unde s and he co ec dis ibu ion ollowed by he da a collec ed, as i is inco ec o assume ha all
da a ollows a i s sigh a no mal dis ibu ion. As he e we e suspicions ha ou dependen
a iable, inal g ade/sco e, did no ollow a no mal dis ibu ion, a Kolmogo o -Smi no es was used
o e i y his hypo hesis and he esul s showed ha , in ac , he e is a high s a is ical p oo ha ou
dependen a iable does no ollow a no mal dis ibu ion as he null hypo hesis was ejec ed wi h a
13
signi icance o 1%. Fu he mo e, i was obse ed ha he a iable dis ibu ion his og am is
asymme ical, and as so, i suppo s he p e ious s a emen ha he a iable doesn’ ollow a no mal
dis ibu ion.
Tes o No mali y
Tes
S a is ic
P-Value
Kolmogo o -Smi no
D = D = 0,099641
P <D <0,0100
Table 4.1 – No mali y Tes
As we ejec ed he hypo hesis ha s uden s’ g ade ollows a no mal dis ibu ion, o assu e he e is
no iola ion o s a is ical es s’ assump ions, he choice was o analyse he da a h ough non-
pa ame ic es s. In his si ua ion, he Mann-Whi ney es was used o compa e wo independen
samples, he K uskal-Wallis o compa e mo e han wo independen samples, bu also he a iances
es , called he Cono e es ha measu es i wo o mo e samples ha e he same a iance, i.e., he
same asymme y o inal classi ica ions, W.J. Cono e said ha “nonpa ame ic me hods use
app oxima e solu ions o exac p oblems, while pa ame ic me hods use exac solu ions o
app oxima e p oblems” (K. M. Ramachand an & Tsokos, 2015).
To begin he explana o y da a analysis, we can s a by analyzing s uden s’ cha ac e is ics on his
s udy: s uden s om Po uguese public high-school, in he yea s 2014/2015 o 10 h, 11 h and 12 h
g ades, e alua ed and a ending he 21 cou ses whe e he na ional exams ake place, as showed on
able 4.2.
As i can be obse ed, mo e han hal he s uden s is emale (55.9%) and hose whose age is
comp ised be ween 16 and 18 yea s old a e he majo i y (81.5%). Focusing on he numbe s uden s
ep o e yea s, i is possible o obse e ha om he ull s uden ’s sample, he majo i y has ne e
ep o ed be o e 10 h (9.4%), 11 h (32.5%) and 12 h g ades (28.5%). The nex highes sco e belongs o
s uden s om 10 h, 11 h and 12 h g ades ha ep o ed once (2.6%, 11,8% and 12.6%, espec i ely). I
is also possible o conclude ha he s uden s who p esen he lowes a e in his a iable a e
egis e ed in he 10 h g ade s uden s who ep o ed wice o mo e (0.2%). Conce ning o na ionali y,
97.2% o he s uden s a e Po uguese and only 2.8% come om o he coun y. Rega ding compu e
Figu e 4.1 – Final Classi ica ion do no ollow a
no mal dis ibu ion
14
access, 71.2% o he s uden s ha e access o compu e s, while 28.8% con i m no ha e compu e
access. In e ne access has e y simila esul s, wi h 68.6% o s uden s claiming o ha e access o he
In e ne , compa ed o 31.4% who said hey did no .
Rega ding he esul s ob ained in he Mann-Whi ney es , i is e i ied ha he e is s a is ical
e idence o he di e ence on academic achie emen be ween gende , in o he wo ds, emale
s uden s end o ob ain be e esul s on mean alues o academic esul s (13.38) han hei male
colleagues (12.90). The esul s ob ained in he Cono e es indica e ha he e a e g ea e
asymme ies in he inal a e age sco e o he male s uden s, his means ha he sco es ob ained by
male s uden s a e mo e i egula ( emales wi h 2.98 and males wi h 3.04), he e a e a g ea numbe
o s uden s who ob ained good esul s and a he same ime o he s who go poo e esul s.
Conce ning he age a iable, i is obse able, h ough K uskal-Wallis es ha a leas one o he
classes ends o achie e s a is ical di e en alues han a leas one o he o he classes.
Ne e heless, h ough he Cono e , i.e., Va iance es , i is possible o conclude ha he lowes
asymme ies in s uden s’ academic achie emen a e eco ded in he class o s uden s be ween 19
and 21 yea s old (2.67), ollowed by s uden s be ween 16 and 18 yea s old (2.99).
Rega ding he esul s ob ained in a iable N_Rep o , which de e mine he numbe o ep o e yea s
ha he s uden has un il he cu en yea and academic pe iod (2014/2015), K uskal-Wallis es
indica es ha wi hin he exis ing classes o each yea he e is a leas one ha egis e s a highe
alue, and i is clea ly suspec ed ha a e he classes o he s uden s ha belong o 10 h (13.18), 11 h
(13.77) and 12 h (13.64) ha ha e ne e ep o ed a yea be o e. On he o he hand, he Cono e es
indica es ha on he 10 h yea he e a e g ea e disc epancies on he esul s o s uden s who ha e
ne e ep o ed (3.15), on he 11 h on he s uden s who ep o ed mo e han wice (3.72) and on he
12 h again he ca ego y o s uden s who ep o ed mo e han wice (3.51).
The s a is ics es s o he Mann-Whi ney es , on a iable na ionali y, show us ha he e a e
di e ences on s uden 's mean alues o academic esul s, wi h na i e Po uguese s uden s (13.19)
ending o ob ain be e esul s in compa ison o s uden s o o he na ionali ies (12.37). Acco ding o
Cono e esul s, bo h Po uguese and immig an s ha e equal asymme ies in hei school esul s
(3.01).
Rega ding s uden s wi h compu e and In e ne access, i is possible o conclude ha he e a e
di e ences in hei academic achie emen when compa ing o he ones ha do no ha e access o
bo h echnologies. The Mann-Whi ney esul s shows ha s uden s who ha e access o compu e
(13.17) and In e ne (13.20) end o ha e be e mean alues o academic esul s. Howe e , on he
Cono e es we obse e g ea e asymme ies in he esul s o he inal a e age sco e, ob ained
among s uden s who ha e access o compu e o in e ne (3.08).
15
Va iables
n
%
Mean
SD
Mann-Whi ney
/K uskal-Wallis (k)
Cono e
Va iance Tes
Gende
Female
62.174
55.9%
13.38
2.98
-2548.4738***1
-2.5849***
Male
49.128
44.1%
12.90
3.04
Age (k)
[0-16[
153
0.2%
13.98
3.46
14072.9362***
3777.8620***
[16-18[
90.682
81.5%
13.36
2.99
[19-21[
19.731
17.7%
11.45
2.67
]>=21]
736
0.7%
11.74
3.66
N_Rep o by yea
10 h,0 ep
10.475
9.4%
13.18
3.15
4050.3054***
372.5845***
10 h,1 ep
2.870
2.6%
11.40
2.93
10 h,2 eps
187
0.2%
10.82
2.91
10 h, +2 eps
209
0.2%
12.08
2.95
11 h,0 ep
36.124
32.5%
13.77
2.89
16579.2848***
2644.0207***
11 h,1 ep
13.116
11.8%
11.95
2.65
11 h,2 eps
976
0.9%
11.09
2.64
11 h, +2 eps
444
0.4%
11.08
3.72
12 h,0 ep
31.725
28.5%
13.64
2.88
9025.3129***
1871.5462***
12 h,1 ep
13.990
12.6%
11.61
2.67
12 h,2 eps
785
0.7%
10.93
2.86
12 h, +2 eps
401
0.4%
11.83
3.51
Na ionali y
Po ugal
108.134
97.2%
13.19
3.01
730.1785***
-6.0427***
O he
3.168
2.8%
12.37
3.01
Compu e
0
32.110
28.8%
13.14
2.84
44.3024***
-37.7989***
1
79.192
71.2%
13.17
3.08
In e ne
0
34.922
31.4%
13.10
2.86
144.2750***
-37.2340***
1
76.380
68.6%
13.20
3.08
Table 4.2 – S uden s’ Cha ac e is ics
Rega ding he desc ip i e analysis o he pa en s’ socioeconomic cha ac e is ics, p esen ed on able
4.3, i is e i ied ha mos s uden s a e no Bene icia y_SASE (73.5%), exis ing a simila dis ibu ion
among he s uden s wi h Bene icia y_SASE in le els 1 (13.7%) and le el 2 (12.8%). The K uskal-Wallis
es indica es ha , in ac , he e a e di e ences in s uden s’ academic achie emen when compa ing
he esul s o mean alues o academic esul s on he h ee le els men ioned abo e. Acco ding o he
esul s i is suspec ed ha s uden s wi h no suppo om SASE a e he ones who ob ained be e
esul s on mean alues o academic esul s (13.33). On he o he hand, he Cono e es shows ha
1
Fo a signi icance le el o 1%, we ejec he null hypo hesis (p- alue <0.0001)
22
powe is N_En ollmen s, he numbe o p e ious en ollmen s each s uden did on a ce ain academic
yea (10 h ,11 h and 12 h g ades). con a y o wha could be expec ed, he s uden s wi h wo o mo e
en ollmen s ha e a lowe p obabili y o ep o ing a cou se le el (13%) when compa ing o hose
wi h jus one en ollmen (22%). Rega ding o he s uden s ha en olled he quali a i e cou ses and
a e om he 11 h o 12 h g ade, he mos disc imina ing a iable is N_Rep o , whe e s uden s ha
ep o e a leas one yea be o e ha e a highe p obabili y o ep o ing a cou se le el (8%) in
compa ison wi h hose ha ha e ne e ep o e a yea be o e (2%). To wha conce ns s uden s ha
had a leas ep o ed a yea be o e, he age o he s uden s i is once again he mos explainable
a iable, ha ing he s uden s wi h mo e han 20 yea s old, a highe p obabili y o ep o ing a cou se
le el (13%) when compa ing o s uden s unde 20 yea s old (7%).

23
3
3
0 – S uden s wi h posi i e app o al a e; 1 – S uden s wi h ep o e a e
Figu e 4.3 – Decision T ee o academic achie emen a yea le el (Model 2)
24
To explain academic achie emen a yea le el, which ep esen s he ep o e a e in he academic
yea 2014/2015, 111302 obse a ions we e analyzed, and ou model indica es ha 81% o he
s uden s show a posi i e app o al a e and 18% while he o he s uden s p esen a ep o e a e o
19%. He e he a iable b inging he highes ele ance o his model is Yea . The ac ha his is he
mos impo an a iable o unde s and i he s uden has a posi i e a e o no a yea le el, will
immedia ely ell us ha he i s hing he model does is di ide he ee in o wo se s, ha ing he
s uden s on 11 h o 10 h g ade a lowe ep o e a e (10%) when compa ed o s uden s en olled on
12 h g ade (30%). Rega ding s uden s on 10 h o 11 h g ade we can obse e ha he e is an a e age
app o al a e o 90% when compa ed wi h he 10% a e age ep o e a e. The a iable ha p esen s
he highes impac on his g oup o s uden s on 11 h o 10 h g ade is N_En ollmen s. He e we can see
ha s uden s wi h wo o mo e en ollmen s ha e a much highe p obabili y o ep o ing a yea le el
(31%) when compa ing o hose wi h jus one en ollmen (8%). Rega ding s uden s wi h wo o mo e
en ollmen s he a iable ha be e jus i ies hei pe o mance is N_Rep o . He e he wo anges
o med a e s uden s ha ha e ne e ep o e be o e and s uden s ha ep o e a leas one yea
be o e, ha ing he g oup o s uden s ha ne e ep o e be o e a highe p obabili y o ep o ing a
yea le el (67%), compa ing wi h hose ha ha e ep o e a leas one yea be o e (21%). Like he
p e ious model, such an ou come migh no be so expec ed o happen.
Secondly, we ha e he se o s uden s en olled on 12 h g ade. He e we ind ou ha 70% o hem
p esen an a e age app o al a e in he academical yea 2014/2015, while 30% an a e age ep o e
a e. The a iable wi h mo e impac on hese esul s is Age. In his case, he e was a spli o med by
s uden s unde 19 yea s old and s uden s wi h 19 yea s old o mo e. The g oup o s uden s wi h less
han 19 yea s showed a much lowe p obabili y o ep o ing a yea le el (23%) compa ing o
s uden s wi h 19 yea s o mo e (45%). To wha espec s on s uden s ha ha e less han 19 yea s old
he a iable mos ly impac ing he esul s is N_En ollmen s. He e he wo g oups o med a e s uden s
wi h one en ollmen and s uden s wi h wo o mo e en ollmen s. The ones wi h mo e han wo
en ollmen s a much highe p obabili y o ep o ing a yea le el (73%) when compa ed wi h hei
pee s wi h only one en ollmen (20%). Fo he ange o s uden s on he 12 h g ade, unde 19 yea s
old, he a iable ha had a highe explana o y powe is N_Rep o . He e, s uden s ha ep o ed a
leas wo yea s be o e, ha e a highe p obabili y o ep o ing a yea le el (51%) compa ing o hose
wi h less han wo ep o ed yea s (42%).
Un il his poin ou ocus has shed some ligh on he an eceden s o academic achie emen . Fo his
pu pose, we’ e de eloped models wi h cha ac e is ics ha acili a e his objec i e. In o he wo ds,
he pa ame e s in he ees ha e been se up no wi h he goal o minimizing p edic ion e o , bu
a he o gene a e simple and easy o in e p e model ees. I ou goal we e o maximize p edic ion
pe o mance, e en a he expenses o in e p e a ion (i.e., clea ly unde s and he p edic ed le el o
academic achie emen based on simple ules), he models would be es ima ed di e en ly. To ha e
an idea o o wha ex en could we p edic academic achie emen wi h he da a we ha e a ailable,
we ha e es ima ed se e al al e na i e and mo e complex models wi h his speci ic pu pose.
We ha e es ima ed decision ees and g adien boos ees. Con a ily o wha we ha e done
p e iously, we ha e no limi ed he numbe o pa en nodes o wo, i.e., ees we e no necessa ily
bina y. Mo eo e , we allow ees o g ow behind h ee le els. Addi ionally, di e en aining
algo i hms (e.g., CHAID and CRT) we e employed as well as di e en e o measu es. We ha e hen
selec he bes models (one o cou se-le el and o he o yea -le el) o he se e al al e na i es.
25
Al hough we will no in e p e hese model ees ( he models a e in Appendix), ou esul s would be
he ollowing:
Figu e 4.5 – Cumula i e Li and Cumula i e Cap u ed Response o Model 2
Figu e 4.4 – Cumula i e Li and Cumula i e Cap u ed Response o Model 1
26
5. DISCUSSION
5.1. DISCUSSION OF FINDINGS
To ge a be e unde s anding o academic achie emen a cou se le el, we mus i s ealize ha he
a iable ha be e explains alone academic achie emen is he cou se i sel . In o he wo ds, he
speci ic cou se is he mos impo an an eceden o academic achie emen a cou se le el. Hence,
di e en cou ses ha e di e en an eceden s, a leas in impo ance. In his case, gi en ha he ee
is bina y, he bes op ion is o isola e quan i a i e and quali a i e cou ses and o es ima e kind o
“di e en models” o each one and de ine di e en an eceden s o each ype o cou ses ins ead o
ying o i a model o all wi hou dis inc ion. So, compa ing he wo ype o cou ses, quan i a i e
cou ses ha e a highe ep o e a e compa ing o quali a i e cou ses.
Conce ning he knowledge o academic achie emen a yea le el, he i s conclusion made, o
unde s and i he s uden will ep o e o no in 2014/2015, was ha he mos impo an hing o
know is he Yea in which he s uden is en olled. This means ha i is no so e icien ha ing a
possible good explainable model o all cou ses as much as, om he s a , o spli hem in wo
g oups: s uden s om 10 h o 11 h g ade and s uden s om 12 h g ade.
A e he esul s o non-pa ame ic es s and decision ee, i is possible o conclude ha he
ollowing hypo hesis, p esen ed be o e a e e i ied: H1, H2, H3, H4, H5, H6, H7, H9, H10, as hey a e
all a e s a is ically signi ican o s uden s’ academic achie emen . Fo he hypo hesis H8 we could no
in e any conclusions as we did no collec su icien da a.
Rega ding hypo hesis H1 we can conclude ha s uden s om emale gende ob ain be e esul s
han hei pee s om he opposi e gende . This conclusion is ein o ced when we ind ha emale
s uden s ou pe o m male s uden s in almos all six co e cou ses om he seconda y deg ee
(p esen ed in able 4), being he only excep ion Fo eign Language – English. Fo hypo hesis H2, he
esul s show ha Po uguese na i e s uden s p esen highe academic esul s, compa ing o o eign
s uden s. In hypo hesis H3 and H4 we e i y ha s uden s wi h access o bo h compu e and in e ne
can achie e be e esul s. Howe e , he esul s om he decision ees show ha o quan i a i e
cou ses, he use o compu e should be mo e mode a e since hey ha e sligh ly lowe app o al a es.
This ein o ces he idea ha , being Po ugal a de eloped coun y, we should in es men in p o iding
be e condi ions so ha ou s uden s could ha e access o such ools, pa icula ly in schools whe e
s uden s spend much o hei ime. Fo hypo hesis H5 he decision ee esul s con i m ha he
numbe o yea s ha he s uden ep o ed be o e is an impo an ac o o explain academic
achie emen , mainly o quan i a i e cou ses and o s uden s en olled on 11 h o 12 h g ade om
quali a i e cou ses i we a e e e ing o he academic achie emen cou se le el. This is also a key
ac o o explain academic achie emen a yea le el, especially o s uden s en olled on 10 h o 11 h
g ade wi h wo o mo e en ollmen s and o s uden s on 12 h wi h 19 yea s old o mo e. This ells us
ha , o example, eache s should be mo e obse an o he his o ic o ep o e a es. Based on
hese esul s, i would be in e es ing o unde s and which a e he phycological implica ions a ep o e
has on he s uden s academical pa h. Fo H6 and H7, ega ding SASE suppo and amily inancial
suppo , we can conclude ha s uden s ha ecei e one o bo h suppo s, ha e a wo se academic
achie emen . Howe e , being s uden s egis e ed in public ins i u ions, wi h equal lea ning
oppo uni ies, his si ua ion should no occu . Fo hypo hesis H9 and H10, as i was al eady
27
men ioned on li e a u e e iew, ou s udy ein o ces, e en mo e, he impo ance o educing he
numbe o s uden s pe class, and consequen ly he numbe o s uden s pe school, as ha would
allow a much be e accompanimen by eache s o he needs o each o hei s uden s.
5.2. PRACTICAL IMPLICATIONS
Se e al p ac ical implica ions can be d awn om his s udy. Fi s his s udy ein o ces he idea o he
impo ance on in es ing on a socie y and especially on an educa ion wi h be e digi al and
echnological ne wo ks, which can be s imula ed by inancing schools and classes wi h compu e s,
echnical manuals o IT suppo , school p og ams o e en p oposing p og amming classes as a
manda o y o e en op ional as we could wi hd aw om his pape ha s uden s wi h compu e and
in e ne access a e capable o be e academical esul s. This enhances he u gency o p omo ing and
suppo ing inc easingly science and inno a ion p og ams.
Secondly, i is equally impo an o con inue o be and aid s uden s wi h less income, as we ha e
ound and concluded om his wo k ha , hese con inue o be he g oup o s uden s wi h lowe
academic achie emen esul s. This suppo may s a , o example, by subsidizing he dis ibu ion o
ma e ials and school meals.
Thi dly, i con inues o be c ucial o in es on he educ ion o he numbe o s uden s pe school and
consequen ly he numbe o s uden s pe class. As so, i would be much easie o eache s o engage
and be e unde s and s uden s cons ains and needs. In addi ion, an e o should be made o
eco e and c ea e be e condi ions and com o in public Po uguese schools as hese a e he
places whe e s uden s spend mos o hei ime. These implica ions a e suppo ed by he ac ha ,
a e eaching ou conclusions and esul s, classes be ween 11 and 30 s uden s a e hose wi h be e
academic achie emen
Las ly, ou indings poin ou and ein o ce he impo ance o knowing s uden s’ schola backg ound
o wha e e s o he numbe o ep o es o p e ious good pe o mance, as he pa h one akes may
possibly be a key ac o o explain hei academic achie emen .
5.3. THEORETICAL IMPLICATIONS
Conce ning he di e en heo e ical implica ions, i is i s impo an o no e ha da a mining
echniques p o ed o be yield good esul s, especially he use o non-pa ame ic me hods as decision
ees gi en he cha ac e is ics o ou da a. This esul makes us hink ha da a mining me hods a e
an eligible and much alid al e na i e o he classical econome ic echniques used on mos o he
s udies conduc ed on he a ea o s uden s’ academic achie emen . These a e echniques ha
p o ide good esul s as hey a e speci ic ools o handle la ge quan i ies o da a and wi h highly
de ailed analysis capable o answe ing o di e en e o s ha big da a bases migh con ain (ou lie s,
missing alues, a iables ans o ma ion, s a is ical analysis). Wi h his in mind, i is ecommendable
ha mo e esea che s use hese new da a analysis echniques.
Secondly he non-pa ame ic es s appea o be equally eliable and a good al e na i e o pa ame ic
es s as hey can handle a iables ha do no ollow a no mal s a is ical dis ibu ion, which in mos
cases limi s mos o classical app oaches and echniques. Besides, hey a e es s easy o implemen
p o iding good esul s.

28
6. CONCLUSIONS
Unde s anding he ac o s ha ha e g ea e impac on academic achie emen is a opic a om
being esol ed, in ac , he e is s ill a lo o imp o e. Howe e , he bigges su plus o his s udy was
he possibili y o wo king wi h da a ha co ec ly ansla ed he eali y o whe e we s and in e ms o
he educa ional le el a seconda y deg ee, al hough we only ocused on academic yea 2014/2015.
Ou indings sugges ha he e a e s ill di e ences and gaps on academic achie emen among
emale and male gende s, as emale s uden s ob ain be e esul s on academic achie emen . We
can also poin ou ha access o compu e s and In e ne , when well used o school pu poses a e a
powe ul mean o help s uden s achie e be e esul s. S uden s coming om less weal hy
households ob ain lowe schola pe o mances and is c ucial o u gen ly ac on his opic ha s ill
s ains ou educa ional pa adigm. Finally, we can assess ha he s uden ep o e backg ound s ill has
a g ea emo ional weigh on his academic achie emen . The cu en esea ch p o ides a d illdown
analysis, which allows us o disco e indings no only a cou se le el bu also a yea le el.
29
7. LIMITATIONS AND RECOMMENDATIONS FOR FUTURE WORKS
Despi e ou bes e o s, some limi a ions mus be acknowledged. The i s one is ega ding he da a
quali y. Al hough MISI da abase comp ises e e y s uden en olled in he Po uguese high school
sys em, being a sou ce o emendous po en ial in educa ion da a, some u he de elopmen s a e
needed in he way da a is eco ded and s o ed. In ac , da a p e-p ocessing ook a e y meaning ul
pa o he e o s conduc ed in his s udy. Missing alues and da a inconsis ency a e aspec s o
imp o e. Secondly, he s udy is in espec o a speci ic poin in ime, i.e., da a used is c oss-sec ional.
In he u u e, i would be in e es ing o do analysis on academic achie emen o mul iple poin s in
ime, whe e each s uden is “ acked” h ough his/he high school expe ience. This app oach would
u he shed ligh on academic achie emen an eceden s. In hi d place, as we ha e used seconda y
da a, we couldn’ include o he po en ial an eceden s o academic achie emen , such as eache s’
cha ac e is ics o e en ( ex ual) no es on s uden s’ beha io by assidui y. Finally, we also
acknowledge some limi a ions in e ms o me hods employed, which is ela ed wi h he p e ious
limi a ion. Ha e we had he oppo uni y o include addi ional a iables, speci ically in e al ones, and
he me hods used would be di e en . We p obably ha e used neu al ne wo ks and eg ession
analysis o imp o e ou p edic ions, and explana ion, on academic achie emen . Mo eo e , he high
sample size o ou da a implies ha non-pa ame ic es s will be much mo e likely o ejec he null
hypo heses, as men ioned ea lie .
30
8. BIBLIOGRAPHY
Abosede, S., & Akin ola, O. (2016). Mo he s ’ Employmen , Ma i al S a us and Educa ional Le el on
S uden s ’ Academic Achie emen in Business S udies. Mul idisciplina y Resea ch, 4(2), 159–
165.
A chibald, S. (2006). Na owing in on Educa ional Resou ces Tha Do A ec S uden Achie emen .
Peabody Jou nal o Educa ion, 81(4), 23–42. h ps://doi.o g/10.1207/s15327930pje8104_2
B. Spina h. (2012). Academic Achie emen . In V. S. Ramachand an (Ed.), Encyclopedia o Human
Beha io (Second Edi, pp. 1–8).
Ba na d, W. (2004). Pa en in ol emen in elemen a y school and educa ional a ainmen . Child en &
You h Se ices Re iew, 26(1), 39. h ps://doi.o g/10.1016/j.childyou h.2003.11.002
Ba ne , R., Glass, J. C., Snowdon, R., & S inge , K. (2002). Size, Pe o mance and E ec i eness: Cos -
Cons ained Measu es o Bes -P ac ice Pe o mance and Seconda y-School Size. Educa ion
Economics, 10(3), 291–311. h ps://doi.o g/10.1080/09645290210127516
Ba o, R., & Lee, J.-W. (2001). In e na ional da a on educa ional a ainmen : upda es and
implica ions. Ox o d Economic Pape s, 53(3), 541–563. h ps://doi.o g/10.1093/oep/53.3.541
Be y, M. J. a., & Lino , G. S. (2011). Da a mining echniques: o ma ke ing, sales, and cus ome
ela ionship managemen . Wiley Publising (Thi d). Re ie ed om
h p://po al.acm.o g/ci a ion.c m?id=983642
Bo zekowski, D., & Robinson, T. (2012). The Remo e, he Mouse, and he No. 2 Pencil. A chi es o
Pedia ics and Adolescen Medicine, 159(2), 607–613.
h ps://doi.o g/10.1001/a chpedi.159.7.607
Boswo h, R. (2014). Class size, class composi ion, and he dis ibu ion o s uden achie emen .
Educa ion Economics, 22(2), 141–165. h ps://doi.o g/10.1080/09645292.2011.568698
Bou dieu, P. (1973). Cul u al ep oduc ion and social ep oduc ion. Pape s in he Soiology o
Educa ion, 71–112.
Bou dieu, P. (1984). Dis inc ion: A Social C i ique o he Judgmen o Tas e. Camb idge, MA:Ha a d
Uni e si y P ess. h ps://doi.o g/10.1007/s13398-014-0173-7.2
Bowles, S., & Le in, H. (1968). The de e minan s o scholas ic achie emen --An app aisal o some
ecen e idence. The Jou nal o Human Resou ces, 3(1), 3–24. h ps://doi.o g/10.2307/144645
B unne , M., Gogol, K., Sonnlei ne , P., Kelle , U., K auss, S., & P eckel, F. (2013). Gende di e ences
in he mean le el, a iabili y, and p o ile shape o s uden achie emen : Resul s om 41
coun ies. In elligence, 41(5), 378–395. h ps://doi.o g/10.1016/j.in ell.2013.05.009
B unne , M., K auss, S., & Kun e , M. (2008). Gende di e ences in ma hema ics: Does he s o y
need o be ew i en? In elligence, 36(5), 403–421. h ps://doi.o g/10.1016/j.in ell.2007.11.002
Buddin, R., & Zama o, G. (2009). Teache Quali ica ion and S uden Achie emen in u ban
elemen a y School. Rand Educa ion, 1–48.
Cabo , J., De Vi gilio, R., & To lone, R. (2017). Web enginee ing: 17 h in e na ional con e ence, ICWE
2017 Rome, I aly, June 5-8, 2017 p oceedings. In Lec u e No es in Compu e Science (including
subse ies Lec u e No es in A i icial In elligence and Lec u e No es in Bioin o ma ics) (Vol. 10360
31
LNCS). Sp inge . h ps://doi.o g/10.1007/978-3-319-60131-1
Ca o, D., Mcdonald, J., & Willms, J. D. (2009). Socioeconomic S a us and Academic Achie emen
T ajec o ies om Childhood o Adolescence. Canadian Jou nal o Educa ion, 32(3), 558–590.
Chowa, G., Masa, R., Ramos, Y., & Ansong, D. (2015). In e na ional Jou nal o Educa ional
De elopmen How do s uden and school cha ac e is ics in luence you h academic
achie emen in Ghana ? A hie a chical linea modeling o Ghana You hSa e baseline da a.
In e na ional Jou nal o Educa ional De elopmen , 45, 129–140.
h ps://doi.o g/10.1016/j.ijedude .2015.09.009
Clo el e , C., Ladd, H., & Vigdo , J. (2006). Teache -S uden Ma ching and he Assessmen o Teache
E ec i eness. Jou nal o Human Resou ces, 41(Ap il 2005), 778–820.
h ps://doi.o g/10.2307/40057291
Codjoe, H. (2007). The impo ance o home en i onmen and pa en al encou agemen in he
academic achie emen o A ican-Canadian you h. Canadian Jou nal o Educa ion, 30(1), 137–
156. h ps://doi.o g/10.2307/20466629
Coleman, J. (1988). Social Capi al in he C ea ion o Human Capi al. Ame ican Jou nal o Sociology,
94(1988), 95–120. h ps://doi.o g/10.1086/228943
Coleman, J., & Hopkins, J. (1966). Equali y o Educa ional Oppo uni y. U.S. Depa men o Heal h,
Educa ion and Wel a e, 66–675. h ps://doi.o g/10.2307/2091096
C oninge , R., Rice, J., Ra hbun, A., & Nishio, M. (2007). Teache quali ica ions and ea ly lea ning:
E ec s o ce i ica ion, deg ee, and expe ience on i s -g ade s uden achie emen . Economics
o Educa ion Re iew, 26(3), 312–324. h ps://doi.o g/10.1016/j.econedu e .2005.05.008
DGE. (2018). Ação Social Escola | Di eção-Ge al da Educação. Re ie ed May 11, 2018, om
h p://www.dge.mec.p /acao-social-escola
D iessen, G., Smi , F., & Sleege s, P. (2005). Pa en al In ol emen and Educa ional Achie emen .
B i ish Educa ional Resea ch Jou nal, 31(4), 509–532.
h ps://doi.o g/10.1080/01411920500148713
D onke s, J., Van De Velden, R., & Dunne, A. (2012). Why a e mig an s uden s be e o in ce ain
ypes o educa ional sys ems o schools han in o he s? Eu opean Educa ional Resea ch Jou nal,
11(1), 11–44. h ps://doi.o g/10.2304/ee j.2012.11.1.11
Egali e, A., & Kisida, B. (2016). School size and s uden achie emen : a longi udinal analysis. School
E ec i eness and School Imp o emen , 27(3), 1–12.
h ps://doi.o g/10.1080/09243453.2016.1190385
Fan, X., & Chen, M. (2001). Pa en al in ol emen and s uden s’ academic achie emen : A me a-
analysis. Educa ional Psychology Re iew, 13(1), 1–22.
h ps://doi.o g/10.1023/A:1009048817385
Gabine e do Sec e á io de Es ado da Educação. (2017). Despacho no ma i o no1-A/2017. Diá io da
República. Re ie ed om h ps://d e.p /applica ion/con eudo/106436777
Ghaz ini, S., & Khajehpou , M. (2011). Gende di e ences in ac o s a ec ing academic pe o mance
o high school s uden s. P ocedia - Social and Beha io al Sciences, 15, 1040–1045.
h ps://doi.o g/10.1016/j.sbsp o.2011.03.236