pitch your research in statistics 2017 - frisam · - trois campagnes de mesures accompagnées...

Pitch Your Research in Statistics 2017 Friday, October 6th 2017, 1.30 – 5.30pm,

University of Fribourg, Pérolles 90, Fribourg

PROGRAM

13:30-14:00 Welcome coffee

14:00-14:45 Laurent Donzé: Why fuzzy logic in statistics? An introduction

14:45-15:15 Redina Berkachy: Demonstration with R how to analyse data in a

fuzzy perspective

15:15-15:45 Coffee, demos, and poster session

15.45-16.45 Short presentations

Corinne Hager Jörin: The different interpretations of

2×2 contingency tables

Nayla Sokhn: Structure and Dynamics of niche-overlap graphs

Martin Huber: Detecting heterogeneous effects of an experimental

educational intervention based on machine learning

16:45-17:30 Aperitif and discussion

of 4

Abstracts

Session 1 Laurent Donzé: Why fuzzy logic in statistics? An introduction

Abstract:

The fuzzy set theory can help the statistician to model fuzziness embedded in data. It is extremely well suited for applications involving natural language, and for modelling non-stochastic uncertainties. We focus our attention on data for which imprecision, vagueness or fuzziness, are coming from linguistic variables, derived from linguistic questionnaires. First, we define "fuzzy numbers" and explain the main operators of fuzzy logic. Follows a description of the so-called "fuzzy process". We end our talk by the presentation of several applications of fuzzy statistics.

Rédina Berkachy: Demonstration with R: How to analyse complex survey data in a fuzzy perspective?

Abstract:

Considering our interest in studying questionnaires composed by linguistic variables e.g. categorical variables using fuzzy logic and applying it in real life data, we show how to compute and analyse this type of data in a fuzzy perspective. We propose an individual and global evaluations of the so-called "linguistic questionnaires" using the signed distance defuzzification method, and we provide an application of these evaluations on real data coming from a survey of the financial place of Zurich, Switzerland. At last, our most recent research showed that one can extend statistical inference methods in the classical theory to the fuzzy one. We supply as instance an example of hypotheses testing in this context.

of 4

Session 2 Corinne Hager Jörin: The different interpretations of 2×2 contingency tables

Abstract:

Three different examples of 2x2 tables are presented: one based on the total number fixed design, one based on the one margin fixed design and one based on paired data. The corresponding null hypothesis and the statistical test used to analyse these data are explained. Although the 2x2 table has a simple form, the analysis of such data is not obvious because it depends on the design of the study.

Nayla Sokhn: Structure and Dynamics of niche-overlap graphs

Abstract:

Niche-Overlap graphs depict the competition between species in the nature. We explored these networks from three different perspectives: algorithmic, structure and dynamics. We developed a novel algorithm to detect efficiently chordless cycles and showed that these cycles are numerous in niche-overlap graphs. We moreover investigated the factors influencing their presence, pointing out the important role of omnivores in these graphs. By quantifying the competition between species, we revealed the prevalence of weak links in chordless cycles, and their strong effect in reducing the number of chordless cycles when they are removed from the network. We simulated the abundance of species over time using systems of Lotka-Volterra differential equations. We then related the persistence of species to their structural properties inside the networks using the generalized linear model. Results revealed a strong significant influence of the degree and to a lesser extent of the frequency of chordless cycles and of clustering coefficient.

Martin Huber: Detecting heterogeneous effects of an experimental educational intervention based on machine

learning

Abstract: We present an experimental evaluation of the impact of an information leaflet about environmental implications of coffee production on the awareness about such environmental issues as well as self-assessed buying behaviour among high school and university students in Bulgaria. Specifically, we combine our randomized experiment with a machine learning method, namely regression trees, to investigate whether the causal effects of the leaflet are heterogeneous across specific subgroups of students. We find that w.r.t. creating awareness about environmental issues, the leaflet had the largest (and statistically significant) impacts on older students, in particular those going to non-technical universities, while the effects on younger (i.e. high school) students were mostly not statistically significantly different from zero.

of 4

Posters

Title Presented by Modelling neural circuit sensorimotor transformations in

drosophila larvae Lucia de Andres

Quel air respirent les Suisses romands dans leurs habitations

économes en énergie ? Corinne Hager Jörin

Characterization of double positive and double negative

feedback loop Xavier Richard

Impact of organizational and psychographic variables on

perceptions of work-life conflict

Mathias Rossi and

Mélanie Thomet

Indoor air quality at Perollino day nursery Pascale Voirin

Demos Title Presented by

Interactive Web Apps presentation using Shiny and R Nayla Sokhn

Jupiterhub ASAM Group

MOD E L LIN G N E U R A L C IR C U I T SEN S OR I M OTORT R A N S F O R M AT I O N S I N DROSOPH ILA L A R VA ELucía de Andrés1, Christian Mazza1, Walter Senn2 and Simon Sprecher11Department of Biology, University of Fribourg (Switzerland) 2Deparment of Physiology, University of Fribourg (Switzerland)[email protected]

Experimental larval navigation

Behavioural set up

Larvae arepresented with alight input thatcomes from aprojector at 45°

Modelling larval behaviour

No stimuli: Brownian movement

-10 -5 5 10 15

5

10

-4 -2 0 2 4

0.1

0.2

0.3

0.4

1. Gaussiandistribution ofrandom numbers

2. Random walkin 1D. Randomwalks of 10.000steps each

2000 4000 6000 8000 10000

-150

-100

-50

50

100

3. Random walk in2D. Lists of pair ofnumber using twoindependentgaussians

Metropolis, a biased random walk

The metropolis algortithm can be used toasses how often larvae go against the gradient.

According to this model a system can goupwards in energy with a probabilty:

E = Energy, extrinsic part

0.5 1.0 1.5 2.0 2.5 3.0

0.2

0.4

0.6

0.8

1.0

Steepness of the field

Larval “internal energy”, ability to go against the field

T = Temperature, intrinsic part

Biological purpose of this intrinsic randomness?

The safest behaviour is to take risks

ü To find the best minima you have to go up

ü Trade-off: speed-efficiency

ü Randomness is robust, maybe genetically hard-wired

Fields in the real world are more complicated, they have local and global minima

s

Larvae would feel astrong force that woulddirect their movements“downwards” in thefield while still beingquite random

Possible fields:Light, odour, gustation

A random walk can be biased in a field

Larval response to light: important parameters

Steepness of light gradient, intensity

Larval internal energy, “temperature”Length

Intensity

• Elizabeth A. Kane et al., 2013. Sensorimotor structure of drosophila larva phototaxis. Proceedings of the National Analysis of Sciences 110.• Alex C. Keene et al., 2011. Seeing the light: photobehaviour in fruit fly larvae. Trends in Neuroscience 35. Bibliography

Important navigation parameters

NI=-1NI≈-0.2

How “efficient” are larval navigations?

Larvae consistently avoid light, yet theirmovements seem to be very inefficient for thispurpose, unexpectedly stochastic

No

stim

uli

Ligh

tstim

uli

Larvae have a better navigation indexwhen the steepness of the gradient andthe intensities are higher.

Larval “internal energy” (T) can beestimated from their navigation indexusing the Metropolis algorithm

g1 g3 g6 g9

g1

g3g5 g6

g7

g9

The filtersnamed as g1-g9are all linear andhave increasingintensity andincreasingsteepness.

Larvae are willing to take more risks whenthe slope and the intensities are higher.

g9g6

g3g1

g1g3

g5

g6 g7

g9

QUEL AIR RESPIRENT LES SUISSES ROMANDS DANS LEURS HABITATIONS ÉCONOMES EN ÉNERGIE ? Mesqualair, un projet Ra&D collaboratif pour y répondreJ. Goyette Pernot1, C. Hager Jörin2, H. Niculita Hirzel3 et V. Perret4

[email protected] - Les mesures prises en faveur des économies d’énergie dans le bâtiment ne risquent-t-elles pas d’induire une détérioration du cadre bâti, neuf ourénové, au détriment de la santé des occupants ?Méthode - Trois campagnes de mesures accompagnées d’un questionnaire détaillé, se sont succédées entre 2013 et 2016 pour répondre à cette question. Radon,composés organiques volatils (COV) et moisissures ont été quantifiés et identifiés dans un grand nombre de villas. Deux campagnes de mesures ont porté entre2013 et 2015 sur la mesure officielle du radon effectuée dans une pièce de vie du bâtiment durant 3 mois dans 650 bâtiments. Parmi ceux-ci 214 étaient desbâtiments neufs ou rénovés ayant obtenu le label MINERGIE® et les 436 autres ont bénéficié du Programme Bâtiment dans le cadre d’un assainissementénergétique partiel ou global. Entre septembre 2015 et février 2016, la troisième campagne s’est focalisée sur la mesure des polluants chimiques à savoir lescomposés organiques volatiles (COV) et les aldéhydes (ALD) et des bio-contaminants dans un peu moins de 200 habitations parmi les bâtiments déjàcomptabilisés.

Conclusion – Il apparait clairement que les bâtiments construits ou rénovés en respectant des prescriptions adéquates de renouvellement d’air telles quelles lesont avec le label Minergie® sont aussi les bâtiments qui présentent a priori le moins de pathologies. Une grande attention est donc requise lors de la rénovationénergétique de bâtiments traditionnels. Le fait de les isoler sans intégrer de concept de ventilation/aération rime à les enfermer sous une cloche. Ces nouvellesconditions risquent à moyen terme d’induire une détérioration significative des conditions de vie à l’intérieur ainsi que la dégradation du bâtiment lui-même. Desrecommandations de bonnes pratiques sont adressées tant aux professionnels de la construction qu’aux occupants de ces habitations. La rénovation du parcimmobilier, enjeux majeur des années à venir, nécessite une extrême vigilance de la part de toutes les parties.

•Contact établis par les Services cantonaux de l’énergie ainsi que l’agence Minergie® •Mesures effectuées sur une base

volontaire après inscription •Soumission systématique d’un

questionnaire détaillé de 120 questions adaptées au type d’habitation et aux usages des occupant-e-s (taux de réponse effectif 95%)•Un tiers des habitations

mesurées ont obtenu le label Minergie

•Concentration moyenne de radon mesurée en Suisse romande de 180 Bq/m3 d’air avec médiane à 71, min à 5 et max à 4’284 Bq/m3

•Effet du type de bâtiment hautement significatif (p< 0.001)•Effet hautement significatif de la

zone de risque (p< 0.001) telle que définie par l’OFSP. Forte empreinte de la géologie•Effet de l’âge tous types de

bâtiments confondus et de la présence d’une cave naturelle hautement significatif (p< 0.001)

•73 maisons ont retourné 110 prélèvements de surface – 22 provenant de 15 bâtiments Minergie et 88 venant de 58 logements rénovés•32% des prélèvements de surface venant de bâtiments Minergie et 56% de ceux effectués dans les bâtiments rénovés sont positifs pour le développement de moisissures

•Maisons Minergie moins affectées • La chambre est contaminée dans les

bâtiments rénovés • La cave est le seul endroit dans

lequel tous les types de bâtiments présentent des moisissures

• Identification de 73 substances présentes au moins à une reprise sur les capteurs•COV totaux (eq toluène) compris

entre 25 et 2’292 mg/m3. Environ 92% des cas sont inférieurs à la recommandation de 1’000 mg/m3 de l’OFSP et 12% dépassent le seuil de 750 mg/m3 suggéré par Minergie-Eco•Toluène attendu dans tous les

prélèvements en raison de sa source extérieure. Teneurs usuelles de l’ordre de 20 – 50 mg/m3 dans air urbain pollué par le trafic routier•Formaldéhyde toujours inférieur à

125 mg/m3 recommandé par l’OFSP•D-Limonène présent dans 76% des

cas. Terpène associé à des parfums et fragrances de produits de nettoyage• Influence significative du garage

intégré sur la qai (COV tot et BTEX: 2 à 3 fois supérieurs mais demeurent inférieures aux seuils recommandés par le label GI)

166

24

91

110

149

10924

91

110 149

109

Répartition cantonale des bâtiments selon leur type Distribution de fréquence des concentrations en radonselon le type de bâtiment

70 maisonsdont 7 Minergie

Dosimètre passif pourla mesure du radon

Boitier exposé à la poussière durant 3 mois

Lingette pour épousseter après 1 semaine

Badge à charbon actif pour la mesure passive des COV

Distribution des habitations sur le territoire Suisse romand selon le type de bâtiment considéré

Situation de fin d’été:•157 bâtiments échantillonnés en septembre

- 25% de bâtiments neufs- 75% de bâtiments rénovés

• Esp. dominante (>70%) d’été : Cladosporium•Niveau de contamination similaire entre le

neuf et le rénové et d’une tranche d’âge àl’autre. Source extérieure probable.

Situation d’hiver:•160 bâtiments échantillonnés durant 3 mois

d’octobre à décembre- 26% de bâtiments neufs- 74% de bâtiments rénovés

• Esp. dominante (>70%) d’hiver: Pénicillium•Parmi les bâtiments construits entre 1975 et

1999, une plus grande proportion présentent un problème de moisissures. Source intérieure cachée?

Radon - Distribution des concentrations mesurées selon le type de bâtiment et le niveau de risque

Moisissures - Identification d’espèces dominantes dans la poussière époussetée selon le type de bâtiment et la saison

Moisissures – Présence/absence effective de moisissures prélevées sur lame selon le type de bâtiment

COV totaux – Distribution des concentrations mesurées selon les types de bâtiments

2013-2016

1

3

2

4

Lingette pour épousseter après 1 semaine

Boitier exposé à la poussière durant 3 mois

Fin d’été Hiver

Lame pour prélèvement sur surface

0%

5%

10%

15%

20%

25%

30%

Aspergillus

Cladosporium

Penicillium

0%

5%

10%

15%

20%

25%

30%

20

00

-20

15

19

75

-19

99

19

50

-19

74

avan

t 1

95

0

Aspergillus

Cladosporium

Penicillium

CHARACTERIZATION OF DOUBLE POSITIVE AND DOUBLE NEGATIVEFEEDBACK LOOP

{ XAVIER RICHARD, BENOÎT RICHARD, CHRISTIAN MAZZA, JAN RELOF VAN DER MEER } UNIVERSITY OF FRIBOURG

INTRODUCTIONAccording to [1], a feedback "refers to a situationin which two (or more) dynamical systems areconnected together such that each system influ-ences the other". A dynamical system is a systemwhose behavior change over time. A feedback ispositive when a perturbation of a system increasethe response of the other. In contrast, a feedbackis negative when a perturbation of a system de-crease the response of the other. Double positiveand double negative feedbacks loops are the sim-plest systems in which two stable states can arise.

STABLE STEADY STATESThe steady states are the solutions of the systemwhen the equilibrium is reached:{

fp,n(A,B) = 0

gp,n(A,B) = 0(3)

A steady state s = (As, Bs) is a stable node ifall the eigenvalues of Js, the Jacobian matrix ofthe system above evaluated at s = (As, Bs) havenegative real part. If all the eigenvalues of theJacobian matrix are positive, the steady state iscalled an unstable node. A system with two sta-ble steady states is called bistable.

REFERENCES

[1] Karl Johan Aström and Richard M. Murray. Feed-back Systems. Princeton University Press, 2th edi-tion, 2009.

[2] D. Gonze and M. Kaufman. Theory of non-lineardynamical systems. University Lecture, 2015.

FUTURE RESEARCHThe next step of this research is to characterize double positive and double negative feedbacks whenan extra feedback (either positive or negative) is added to one of the species. The systems of equationsbecome much more complex and news tools such as interval analysis are needed in order to characterizethose systems. An other further work is the characterization of three interacting species.

CONTACT INFORMATIONEmail [email protected] Department of Mathematics

Chemin du Musée 23CH-1700 Fribourg

Phone +4126 300 9201

PHASE PLANE ANALYSISIn order to see graphically the dynamic of the sys-tem, a phase plane analysis can be perform forgiven parameters c1 and c2 in equations (1) and(2). When the arrows are going in the direction ofan intersection point, this point is a stable node.When the arrows are going away from an inter-section point, this point is an unstable node.

0 2 4 6 8A (a.u)

0

2

4

6

B (

a.u

)

7B2

1 +B2 A= 0

5A2

1 +A2 B= 0

0 2 4 6 8A (a.u)

71 +B2 A= 0

51 +A2 B= 0

Figure 1: Phase plane analysis of a double posi-tive feedback (left) and of a double negative feedback(right). The parameters are in both cases c1 = 7 andc2 = 5. The two stable curves are defined by eq. (3).

The next step is to determine for which parameterc1 and c2 those systems are bistable.

BISTABILITY RANGE

0 1 2A (a.u)

0

1

2

B (a

.u)

0 1 2A (a.u)

0 1 2A (a.u)

Crossing pointTangency point

Figure 2: Change of the stable curves when the param-eters goes from c1 = c2 = 1.8 (left, one intersection), toc1 = c2 = 2.2 (right, three intersections). A tangencypoint is encountered for c1 = c2 = 2 (center).

Having two or more intersections is a sufficientcondition for bistability. The fact that deformingthe stable curves to change the number of inter-section entail the creation of a tangency point al-lows to find the boundary of the bistable region.

0 5 10 15 20c1 (a.u)

0

5

10

15

c 2 (a

.u)

bistable configurationanalytic solution

0 5 10 15 20c1 (a.u)

bistable configurationanalytic solution

Figure 3: Bistability range of a double positive feed-back (left) and a double negative feedback (right). Eachdot represents a configuration which has been testednumerically to be bistable and the solid line the ana-lytic solutions corresponding to the boundary betweenbistable and non bistable area.

The boundary is found as a parametric curve. Fora double positive feedback loop it yields{

a(λ) =(1 + λ−2

)√4 (1 + λ2)

−1 − 1

b(λ) = 4λ(3− λ2

)−1

and for a double negative feedback loop{a(λ) =

(1 + λ2

)√(1 + λ2) (3λ2 − 1)

−1

b(λ) = 4λ3(3λ2 − 1

)−1.

STOCHASTIC SIMULATIONThe Gillespie’s algorithm was used to performstochastic simulations in order to confirm thebistability of the system and to see how "deep"each stable node is. A total of 10000 simulationswere ran, with A and B ranging from 0 to 100 asinitial condition.

0 4 8 12 16A (a.u)

0

4

8

12

B (

a.u

)

0 4 8 12 16A (a.u)

density

Figure 4: Stochastic simulations of a double positivefeedback (left) and a double negative feedback (right).The parameters corresponding to the equations (1) and(2) are c1 = 7 and c2 = 5. In each cases, the valueswith the higher density are those corresponding to thesection "phase plane analysis".

In the case of a double positive feedback loop ,the point p = (0, 0) is an absorbing point i.e.if the simulation is running a certain time, bothspecies A and B will be stuck in (0, 0). In the caseof a double negative feedback, there is always aprobability to switch from one stable node to theother.

CONCLUSIONBased on this research, the following conclusionsarise:

• The two systems can be bistable.• The analytic solutions match the numerical

computations.• The stochastic simulations show how

"deep" is each stable node.

MODEL

Double positivefeedback loop

A B

Double negativefeedback loop

A B

A double positive feedback can be expressed bythe following differential equation system:

dA

dt=

c1B2

1 +B2−A = fp(A,B)

dB

dt=

c2A2

1 +A2−B = gp(A,B)

(1)

A double negative feedback can be expressed bythe following differential equation system:

dA

dt=

c11 +B2

−A = fn(A,B)

dB

dt=

c21 +A2

−B = gn(A,B)

(2)

I m p a c t o f o r g a n i z a t i o n a l a n d p s y c h o g ra p h i c v a r i a b l e s o n p e r c e p t i o n s o f w o r k ‐ l i f e c o n f l i c t

Contact:

Mélanie Thomet and Mathias Rossi

Ins tute for Social and Public Innova on

HES‐SO Haute école spécialisée de Suisse occidentale

School of Management Fribourg

[email protected] / [email protected]

+41 (0)26 429 63 96

Purpose

The objec ve of this research is to understand what means are available to an

organiza on to prevent the psychosocial risks arising from a conflict between

work and personal life (which is a combina on of psychosocial risk factors), focu‐

sing on different organiza onal variables (autonomy, job requirements, colleague

support, hierarchy and organiza on), as well as a psychographic variable (the

sense of coherence).

Hypothesis

General hypothesis: There is a percep on of conflict between work and perso‐

nal life, which is more or less present based on the percep on of the work situa‐

on.

Specific hypothesis:

H1: The higher the level of sense of coherence, the lower the percep on of a

conflict between work and personal life.

H2: The higher the level of autonomy within the workplace, the lower the percep‐

on of work‐life conflict.

H3: The higher the percep on of job requirements, the greater the percep on of

work‐life conflict.

H4: The higher the percep on of support from colleagues and employee's hie‐

rarchy, the lower the perceived conflict between work and personal life.

H5: The higher the percep on of organiza onal support, the lower the perceived

conflict between work and personal life.

Methodology

The quan ta ve diagnosis was carried out in a representa ve sample of 60 indi‐

viduals. A total of 44 people par cipated in this study, with a response rate of

73%. The number of surveys handled sta s cally remains of the order of 44.

The quan ta ve survey consisted of 62 ques ons: socio‐demographic ques ons

(age, gender, presence of children and sector) ; Antonovskys Life Orienta on

Ques onnaire (seven‐point seman c differen al scale with two terms (or

phrases) disposed at the extremes) ; Job Content Ques onary (JCQ) of Karasek

(four‐point Likert scale) ; Eisenberger scale (five‐point Likert scales) ; Geurts

scale, named SWING (four‐point Likert scale).

Results (socio‐demographic variables)

Organiza onal variables (perceived autonomy,

perceived task demands, support from colleagues, hierarchy and perceived

organiza on)

Psychographic variable (sense of coherence)

Possible Influence (to check)

Suggested influence

Percep on of a conflict between work and personal

life more or less strong

Sociodemographic variables (age, gender,

presence of children, sector of ac vity)

Sociodemographic variables Significant Trends

Sector *: in this sector the job requirements perceived are higher

> Posi ve impact of work on privacy (in terms of propor ons)

Fisher’s Exact Test : p‐value = 0.066)

Age: over 35 years <Psychological demand (in terms of pro‐

por ons) Fisher’s Exact Test : p‐value = 0.052

Presence of children: no children > Psychological demand (in terms of pro‐

por ons) Fisher’s Exact Test : p‐value = 0.057

Gender: Women

<Social support (in terms of propor ons and medians)

Fisher’s Exact Test : p‐value = 0.054 Wilcoxon test : p‐value = 0.088

Main contribu ons

This study highlighted the impact of certain organiza onal variables (job require‐

ments, social support and organiza onal support) and the sense of coherence in

the percep on of work‐life conflict.

It opens up new perspec ves for concrete recommenda ons for organiza ons.

Main references:

‐ TREMBLAY, Diane‐Gabrielle, 2012. Ar culer emploi et famille: le rôle du sou en

organisa onnel au cœur de trois professions.

‐ GANA, Kamel, LOUREL, Marcel et WAWRZYNIAK, Sophie, 2005. L’Interface « vie

privée ‐ vie au travail » : adapta on et valida on française de l’échelle SWING

(Survey Work ‐ Home Interac on‐Nijmegen).

‐ OWENS, Bradley et WADSWORTH, Lori, 2007. The effects of social support on

work‐family enhancement and work‐family conflict in the public sector.

‐ VAN WASSENHOVE, Wim, 2014. Modèle de Karasek.

‐ DEBRAY, Caroline, [et al. ] 2014. Le lien entre le sen ment de cohérence, le stress

et la santé chez le dirigeant de PME: résultats d’étude.

Results (valida on of hypothesis)

Hypothesis Valida on Sta s cal tests used

General hypothesis Accepted

H1: level of sense of coherence Accepted (for

nega ve effect)

H2: level of autonomy Declined

H3: percep on of job requirements Accepted (for

posi ve effect)

H4: percep on of social support Accepted (for

nega ve effect)

H5: percep on of organiza onal support Accepted (for

posi ve effect)

‐ Shapiro‐Wilk ; Q‐Q‐plot

‐Test on the average of a star ng variable

(test on the right)

‐ Correla on test (Pearson and Spear‐

man)

‐ Par al correla on

Pitch your Research in Sta s cs 2017, Swiss

Sta s cal Society and FRI SAM workshop,

Fribourg, 06.10.2017

Results (valida on of hypotheses)

Variables Descrip ve sta s cs Test of

normality Principal rela ons to be verified Other rela onships to verify

Effet_nega f

Average = 7.91 Median = 7.00 Variance = 9.34

Standard devia on = 3.056

Minimum = 3; Maximum = 18

Min. possible = 0 Max. possible = 24

95% confidence interval = [6.98; 8.84]

No Shapiro‐Wilk: W = 0.916, p‐value = 0.004

Test on the average of a star ng variable (nega ve effect of work on privacy). Test on the right (it is assumed that the average of the police concerning the nega ve effect of work on privacy is greater

than 6). S0 = 4.14

p‐value = 0 The general hypothesis is accepted.

Effet_posi f

Average = 7.34 Median = 8.00

Variance = 10.509 Standard devia on =

3.242 Minimum = 1; Maximum

= 14 Min. possible = 0

Max. possible = 15 95% confidence interval =

[6.36; 8.33]

Yes Shapiro‐Wilk: W = 0.971, p‐value = 0.326

/

Demande_ psychologique




= 29 Min. possible = 9

Max. possible = 36 95% confidence interval =

[22.54; 24.64]


Correla on with the nega ve effect of labor (Spearman): correla on coefficient = 0.228

p‐value = 0.137 Correla on with the posi ve effect of labor (Pearson):

correla on coefficient = 0.372 p‐value = 0.013

H3 hypothesis is accepted for the posi ve effect of labor.

Sou en_social

Average = 9.98 Median = 10.00 Variance = 1.325

Standard devia on = 1.151

Minimum = 7; Maximum = 12

Min. possible = 3 Max. possible = 12


No Shapiro‐Wilk: W = 0.925, p‐value = 0.007

Correla on with the nega ve effect of labor (Spearman): correla on coefficient = ‐0.30

p‐value = 0.048 Correla on with the posi ve effect of labor (Spearman):


H4 hypothesis is accepted for the nega ve effect of labor.

La tude_ decisionnelle




= 94 Min. possible = 42 Max. possible = 96






H2 hypothesis is denied.

Correla on with social support (Spearman): correla on coefficient = 0.411

p‐value = 0.006 Decisional la tude and social support =

posi ve linear rela onship

Sou en_ organisa onnel


Variance = 39,794 Standard devia on =








H4 hypothesis is accepted for the posi ve effect of labor.

Sen ment_ coherence









correla on coefficient = ‐0.068 p‐value = 0.662

H1 hypothesis is accepted for the nega ve effect of labor.

Correla on with psychological demand (Pearson):

correla on coefficient = ‐0.297 p‐value = 0.050

Sense of coherence and psychological de‐mand = nega ve linear rela onship

Par al correla on between sense of co‐

herence, psychological demand and nega ve effect: none

INDOOR AIR QUALITY AT PEROLLINO DAY NURSERYR-SÛR – a collaborative R&D project at the school of engineering and architecture of Fribourg (HEIA//FR)

J. Goyette Pernot, P. Voirin, M. Capitan, Lucy Linder, P. Kuonen, O. Vorlet

The aim of this R‐SÛR project is to demonstrate the feasability and the benefit of keeping a constant check, withremote monitoring, on the indoor air quality (IAQ) in a building accommodating young children (0‐5). The ideais to ensure a 100% healthy environment in the medium term, as this young population is very sensitive and canbe affected in its physical and psychological development. This topic represents a true public health issue.

With this in mind, a major step will be to raise awareness and educate the nursery staff on this issue in thefuture.

Measurements at Perollino Development of a prototype at HEIA//FRThe Amstein&Walthert fireflies device used is a multi‐sensor platformthat measures :

The prototype consists of two sensors boards protected by a box (see picture above) and isconnected to a portable computer by USB. The communication architecture relies on POP‐C++objects and can be generic, independently of the actual sensors used; a driver must be availablefor each sensor.To get and save measurements, we address two databases:• a local database with raw records, that we can question by SQL queries (Structured query

language)• A remote database, hosted by the BBData platform.

This platform offers also an interface to displayeasily the recorded sensors measurements.This is accessible from anywhere on the internetnetwork

https://r‐sur.daplab.ch

Coldest week

Too dry conditions

Pérollino opening

Cleaning and moving in

WHO recommandation for CO2

Increase over 1000 ppm

Source: meteoblue, Fribourg

The comparison of the data collected with our prototype and the flyerflies system is quitepromising; even if there is a shift, the relative behaviour is very close, as for CO2 for instance:

Carbon dioxide, volatile organic compounds, fine particles, radon are part of the indoor pollution sources of of the air breathed all day long.

Up to now, this study has mainly been developed into two parallel parts :• measuring, in collaboration with Amstein &Walthert, the usual indicators with a standard measurement device, to get a first characteristic signature of Perollino nursery air• designing at HEIA‐FR a measurement prototype, as open and flexible as possible, and testing the measurement recording, the data transfer as well as the data analysis

PITCH YOUR RESEARCH IN STATISTICS 2017Friday, October 6 , UNIFR

• Temperature (°C)• Relative humidity (%)• Carbon dioxide CO2 (ppm)• Noise level (dB)

• Light volatile organic compounds (μg/m3)• Total volatile organic compounds (μg/m3)• Fine particles, 0.5 to 1.5 μm (Mpart/m3)

We can observe portions of rawdata directly:

We have implemented several types of sensors, commercially available: calibrated or not, more or less cheap, in order to test what we can measure dynamically. The main advantage, compared to the fireflies platform, is that we can monitor the different sensors, add or remove a sensor according to our wishes or results, thus improving our prototype design with time.

CO2 data recordedwith the fireflies

CO2 data recordedwith the prototype

Further work:Study of sensors calibration – investigation of different sensors – improvement of prototype autonomy and measurement robustness

43.3%

25.1%

19.2%

4.5% 4%2.3% 1.6%

0

10

20

30

40

50

1 2 3 4 5 6 7Dimensions

Perc

enta

ge o

f exp

lain

ed v

aria

nces

Scree plot

We can also summarize the data – 7 variables (see above) observed on 3409 individuals(different timing) – with principal component analysis (PCA). This method consists in building virtual variables to best explain the variance contained in the data. These virtual variables, the components, are linear combinations of the true variables, defined mathematically as the eigen vectors of the correlation matrix. The graph on the right shows the projection of individuals in the plane of components 1 and 2; Arrows are the projections of true variables in the same plane. Their lengths and relative positions show which variables are important and which are correlated.

Further work:Validation of the hypothesis of normality for PCA method.Investigation of classification methods to characterize these time series, with the objective to get identified profiles of indoor air quality in a specific building, in particular at Perollino.

-0.05 0.00 0.05 0.10

-0.0

50.

000.

050.

10

Comp.1

Com

p.2

1

2

3

45

6789101112131415161718

192021222324

2526272829303132333435363738

39404142

434445464748495051525354555657585960616263646566

676869707172737475767778798081828384858687888990919293949596979899100101102103104105

106107

108109110111112113114115116117118119120121122123124125

126

127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160

161162163

164165166167168169170171172173174175176177178179180181182

183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242

243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311

312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356

357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389

390391392393394

395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434

435436437438439440441442443444445446447448

449

450451452453

454455456457

458

459

460461

462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552

553

554555556557558559560

561562

563564

565

566567

568

569

570

571

572

573574

575576

577578

579580

581582583584

585586587588589

590591

592593

594595596

597598

599600601602603604605606607608609610611612613614615616617618619620621622623624625626627628629630631632633634635636637638639640641642643644645646647648649650651

652653654655656657658659660661662663664665666667668669670671672673674675676677678679680681682683684685686687688689690691692693694695696697698699700701702703704705706707708709710711712713714715716717718719720721722723724725726727728729730

731732733734735736737738739740741742743744745746747748749750751752

753754755756757758759760761762763764

765766767768769770

771

772773774775776777778779780781782783784785786787788789790791792793794795796797798799800801802803804805806807808809810811812813814815816817818819820821822823824825826827

828829830831832833

834835836

837

838839840

841842

843844845846

847848

849850

851852853

854

855

856

857858

859

860861862863

864

865

866

867

868

869

870871872

873874

875

876877

878879

880

881 882

883

884885

886887

888889890891892

893894

895

896

897898

899900

901

902903

904905906907

908

909910911912913914915916

917918919920921922923924925926927928929930931932933

934935936937938

939940941942943944945946947948949

950951952953954955956957958959960961962963964965966967968

969970971

972973974

975976

977978979

980

981

982983984

985

986987988

989

990

991

992

993994995

996

997998

999

1000

10011002

1003

100410051006

1007

1008

1009

1010

101110121013

10141015

10161017

1018101910201021

10221023

1024

1025

1026

1027

1028

1029

1030

103110321033

103410351036

10371038

1039

104010411042

10431044

10451046

1047

1048

1049

1050

10511052

10531054

1055

105610571058105910601061106210631064106510661067106810691070107110721073107410751076

1077107810791080108110821083108410851086108710881089109010911092109310941095109610971098109911001101110211031104110511061107110811091110111111121113111411151116

1117

1118

1119112011211122

11231124

1125

1126

1127

11281129

1130

11311132

1133113411351136

11371138113911401141

11421143

1144

1145

114611471148

11491150115111521153

11541155

11561157

11581159

11601161

11621163

1164

1165

1166

1167 11681169

1170

11711172

117311741175

1176

1177

11781179

1180

11811182

1183

1184

1185

1186

1187118811891190119111921193

119411951196119711981199120012011202120312041205120612071208120912101211121212131214121512161217121812191220122112221223122412251226122712281229123012311232123312341235123612371238123912401241124212431244124512461247124812491250125112521253

12541255125612571258125912601261126212631264

12651266126712681269

1270

127112721273

12741275 12761277

1278

1279

12801281

12821283

1284

12851286128712881289

129012911292

1293

129412951296

1297

1298

1299

1300

1301

1302

13031304

1305

1306

13071308

13091310

1311

1312

1313

1314

1315

13161317

1318

13191320

1321

132213231324

1325132613271328

1329

1330

133113321333

133413351336

13371338133913401341134213431344134513461347134813491350

1351135213531354135513561357135813591360136113621363136413651366136713681369137013711372137313741375137613771378137913801381138213831384

138513861387138813891390139113921393

1394

139513961397139813991400140114021403

1404

1405140614071408

140914101411

1412

1413

141414151416

1417

14181419

14201421142214231424142514261427

1428

1429

14301431

143214331434

1435

1436

1437

1438

1439

14401441

1442

1443144414451446

1447144814491450

1451

14521453

1454

1455

14561457

1458

1459

1460

14611462

1463

1464

146514661467

146814691470

1471

1472

1473

1474

1475147614771478

14791480148114821483148414851486148714881489

1490

1491

1492

1493

1494

149514961497

1498

149915001501

1502

15031504150515061507150815091510151115121513151415151516151715181519

1520152115221523152415251526152715281529153015311532153315341535153615371538153915401541154215431544154515461547154815491550155115521553155415551556155715581559156015611562156315641565156615671568156915701571157215731574157515761577157815791580158115821583158415851586158715881589159015911592159315941595159615971598159916001601160216031604

16051606

16071608160916101611

1612161316141615161616171618161916201621

162216231624162516261627162816291630163116321633163416351636163716381639164016411642164316441645164616471648164916501651165216531654165516561657165816591660166116621663166416651666166716681669167016711672167316741675167616771678167916801681168216831684168516861687168816891690169116921693169416951696169716981699170017011702170317041705170617071708170917101711171217131714171517161717171817191720172117221723172417251726172717281729173017311732173317341735173617371738173917401741174217431744174517461747174817491750175117521753175417551756175717581759176017611762176317641765176617671768176917701771177217731774177517761777177817791780178117821783178417851786178717881789179017911792179317941795179617971798179918001801180218031804180518061807180818091810181118121813181418151816181718181819182018211822182318241825

1826

1827182818291830183118321833

18341835183618371838

1839

18401841

1842

184318441845

1846184718481849

1850185118521853

18541855

1856185718581859

186018611862

1863186418651866

18671868

18691870

18711872

1873

18741875

1876

1877

18781879

1880

1881

18821883

18841885

18861887

1888

1889

1890

1891

18921893

18941895

1896

1897

1898

1899

190019011902

1903

1904

19051906

19071908190919101911191219131914

19151916

1917

191819191920

192119221923192419251926192719281929193019311932193319341935193619371938

1939194019411942194319441945

1946194719481949195019511952195319541955195619571958195919601961196219631964196519661967196819691970197119721973197419751976

19771978

1979198019811982

1983

198419851986

198719881989

199019911992

1993

1994

1995

1996

199719981999

20002001

200220032004

20052006200720082009

2010

201120122013

2014

2015 2016

2017

20182019

202020212022

2023

20242025202620272028

202920302031

2032

20332034

2035

2036

2037

2038

2039

204020412042

20432044

204520462047

2048

2049

2050

2051

205220532054

20552056

2057

20582059206020612062

2063

20642065206620672068206920702071207220732074207520762077207820792080208120822083208420852086208720882089

20902091209220932094209520962097

209820992100210121022103210421052106210721082109211021112112

211321142115211621172118211921202121212221232124

212521262127

21282129

213021312132

2133

21342135

2136

21372138

21392140

2141

214221432144

2145

21462147

214821492150

21512152

21532154

215521562157

2158

21592160

2161

2162

21632164

2165

2166

2167

21682169217021712172

21732174

21752176

2177

21782179

21802181

2182

2183

21842185

2186

2187

21882189219021912192

2193219421952196219721982199220022012202

2203220422052206

2207220822092210221122122213221422152216

221722182219222022212222

222322242225222622272228222922302231223222332234223522362237223822392240224122422243224422452246224722482249225022512252225322542255225622572258225922602261226222632264226522662267

226822692270227122722273227422752276

22772278

2279

2280

22812282

22832284

22852286

2287

2288228922902291

22922293

2294

22952296

22972298

2299

23002301

2302

23032304

2305

2306

2307

2308

2309

2310

2311

2312

2313

2314

231523162317

231823192320

2321

23222323

2324

2325

2326

2327

2328

2329

2330233123322333

2334

233523362337

23382339234023412342234323442345234623472348

23492350

23512352

235323542355

235623572358235923602361

2362236323642365236623672368236923702371237223732374237523762377237823792380238123822383238423852386238723882389239023912392239323942395239623972398239924002401240224032404240524062407240824092410

24112412

24132414241524162417241824192420

242124222423

24242425

242624272428

24292430

2431

2432

243324342435

24362437

24382439

24402441

24422443

24442445

244624472448

2449

2450

24512452

2453

245424552456

2457

2458

24592460

24612462

246324642465

2466

2467

2468

2469

2470

24712472

2473

247424752476

24772478

2479

2480

2481248224832484248524862487248824892490

24912492

2493

24942495

24962497

2498

2499

2500

25012502

25032504

2505

25062507250825092510

25112512251325142515251625172518

25192520252125222523252425252526252725282529253025312532

2533253425352536253725382539254025412542254325442545

2546

25472548254925502551255225532554255525562557255825592560256125622563256425652566256725682569257025712572257325742575257625772578257925802581258225832584258525862587258825892590259125922593259425952596259725982599260026012602260326042605260626072608260926102611261226132614261526162617261826192620262126222623262426252626

26272628262926302631263226332634263526362637263826392640264126422643264426452646264726482649265026512652265326542655265626572658265926602661266226632664266526662667266826692670267126722673267426752676267726782679268026812682268326842685268626872688268926902691269226932694269526962697

2698269927002701270227032704270527062707270827092710271127122713271427152716271727182719272027212722272327242725272627272728272927302731273227332734273527362737273827392740274127422743274427452746

2747274827492750275127522753275427552756275727582759

27602761276227632764276527662767276827692770277127722773277427752776277727782779278027812782278327842785278627872788278927902791279227932794279527962797279827992800280128022803280428052806280728082809281028112812281328142815281628172818281928202821282228232824282528262827282828292830

28312832283328342835283628372838283928402841284228432844

284528462847284828492850

28512852

2853

2854

2855

2856

2857

2858

2859 2860

2861

28622863

28642865

2866

2867

2868

2869

2870287128722873

28742875

28762877

2878

287928802881

2882

2883

28842885

2886

2887288828892890

2891

28922893

2894

2895

2896

2897

2898

2899

2900

2901

2902

2903

2904

29052906

2907

2908

2909

291029112912

2913

2914

2915

2916

2917

2918291929202921292229232924

2925

292629272928

2929293029312932293329342935293629372938

29392940294129422943

29442945294629472948

29492950295129522953295429552956295729582959

2960

29612962296329642965296629672968296929702971

29722973

29742975297629772978

29792980298129822983298429852986

298729882989

29902991299229932994

2995

2996

2997

29982999

3000

3001

3002

3003

30043005

30063007

3008

30093010

3011

30123013

30143015

3016

301730183019 3020

3021

3022

3023

3024

3025

30263027

3028

3029

3030

3031

30323033

303430353036

3037

30383039

3040

3041

30423043

3044

30453046

30473048

3049

3050

3051

305230533054

3055

3056

3057

3058

3059

3060

3061

3062

30633064

3065

3066

30673068

3069

3070

30713072

30733074307530763077

30783079308030813082308330843085308630873088308930903091309230933094309530963097309830993100

31013102310331043105

310631073108310931103111311231133114311531163117311831193120312131223123312431253126312731283129313031313132313331343135

3136

31373138

3139

314031413142

31433144

3145

31463147

3148

3149

3150

3151 3152

315331543155

3156

31573158

3159

316031613162

31633164

31653166

3167

3168

3169

31703171

31723173

3174

31753176

3177

3178317931803181

31823183

31843185

3186

3187

31883189

31903191

3192

3193

3194

3195

31963197

3198

31993200

3201

3202

320332043205

32063207

32083209

321032113212321332143215

3216321732183219

32203221322232233224322532263227322832293230323132323233323432353236323732383239324032413242324332443245324632473248324932503251325232533254325532563257325832593260326132623263326432653266326732683269327032713272327332743275327632773278

32793280

3281328232833284

32853286

3287

3288

328932903291

32923293

32943295

3296

3297

329832993300

3301

33023303

3304

33053306

3307

33083309

3310

3311

3312

3313

331433153316

3317

331833193320

3321332233233324

33253326

33273328

3329

3330

33313332

3333

3334

33353336

3337

3338

3339

33403341

334233433344

33453346

3347334833493350

3351335233533354335533563357335833593360

336133623363336433653366336733683369337033713372

33733374337533763377337833793380338133823383338433853386338733883389339033913392339333943395339633973398339934003401340234033404

34053406340734083409

-20 0 20 40

-20

020

40

T

H Bruit

co2

covlcovt

part

Relative magnitude (eigenvalues) of the components

Individuals and variables in the 1‐2 components plane

Pitch Your Research in Statistics

University of Fribourg (Switzerland)Bd de Pérolles 90

CH – 1700 Fribourg (Switzerland)

Interact – Analyze – Communicate

Make interactive web applications forvisualizing dataBring R data analysis to lifeCombine the computational power of R withthe interactivity of the modern web

http://diuf.unifr.ch/asam

@asamunifr


http://diuf.unifr.ch

http://www.unifr.ch


https://www.facebook.com/asamunifr

ASAM Research GroupDepartment of Informatics

University of Fribourg (Switzerland)Bd de Pérolles 90

CH – 1700 Fribourg (Switzerland)

Launched by the ASAM Group

Give students a set of interactive courses inlive environmentsGive students the possibility to create theirown notebooksAccess to standard datascience softwares(R, Python, Julia)Access to numerous data sets

https://asam-group.ch

@asamunifr


http://diuf.unifr.ch

http://www.unifr.ch

https://asam-group.ch

https://www.facebook.com/asamunifr

pitch your research in statistics 2017 - frisam · - trois campagnes de mesures accompagnées...

Documents