Seminar for Statistics

Projects and Reports

HS 2010

Interpretation of microarray data

Student: Meichtry André  [mcdr at]

Client: Anna Egger, Handschin group, Biozentrum Basel


Abstract Problem Description: The aim was to analyse differential gene expression for specific contrasts for two different projects. For the Retina project, the two independent variables were Stimulation with levels light (LI) and no light (DA) and Type with levels knockout (KO) and wildtype (WT). For the Skeletal muscle project, the two independent variables were Virus-Control WT and Virus-Control KO.
Methods: Data were screened and normalised with the robust multiarray average method. Contrasts were constructed with respect to the research questions. The analysis was performed using the linear modelling strategy described by Gordon K. Smyth and which is implemented in the limma package of the Bioconductor project. Especially, moderated t and F-statistics were used that take advantage of the parallel structure of the data. Multiple testing was considered by applying a false discovery rate control strategy. Differential expression was defined by a log fold change larger than 1.2 on the corresponding contrast and a false-discovery-rate-corrected p-value of less than 5% on the gene-specific global moderated F-test.

Full Report [pdf]

CO2 impact on plantlouse-system

Student: Xiaobei Zhou [ xzhou at]
Client:Jeannine Klaiber (ETH Zurich, Institute of Plant, Animal and Agroecosystem Sciences)


Abstract:The goal of this project is to study selected effects of carbon dioxide (CO2) on the Brussels sprouts - cabbage aphid  system. The effects of different levels of CO2 (ambient CO2 and elevated CO2) on selected physical and chemical plant parameters of Brussels sprouts plants grown for an extended period under climatic regimes is analyzed in this report. I also take interest on the acceptance of such plants by winged cabbage aphids and subsequent aphid reproductive performance.

Full Report [pdf]

Staging of Human African Trypanosomiasis patients

Student: Sarah Gerster [gerster at]

Client: Markus Müller (Swiss Institute of Bioinformatics, Proteome Informatics Group); Jean-Charles Sanchez (University of Geneva, Biomedical Proteomics Research Group)


Abstract:The disease Human African Trypanosomiasis (HAT), also called sleeping sickness, occurs in the Sub-Saharan region. The disease is due to the infection by a parasite. There are 2 subspecies of the parasite: gambiense and rhodesiense. The parasite is transmitted to humans by the tsetse fly. Domestic animals serve as hosts. Although not directly affected by the parasite, they serve as pool of parasites. Eliminating the disease physically is not feasible on larger areas. Infected patients can be in two stages. Stage I patients have the parasite only present in their blood and can be treated fairly efficiently with low risk. Stage II patients have the parasite present in their blood and cerebrospinal fluid (CSF). The molecules of the medicine healing stage I patients are too large to cross the blood/CSF membrane and will not heal stage II patients. The standard medicine used to cure stage II patients has a high mortality rate. Hence, before treating a patient with this medicine, one wants to be sure that he/she really is in stage II.

Full Report [pdf]

Effect of AMF on Biomass Production and Root Allocation

Student: Weilian Shi [stoneswl at]

Client: Bernd Felderer (ETH Zurich, Institute for Terrestrial Ecosystems)


Abstract: Phosphorus (P) is one of the most important elements that plants need for growth and biomass production. It is often heterogeneously distributed and limited in soil - an increase of P fertilizer will lead to an increment of biomass production. Plants can uptake P from soil directly from their roots or from fungi called arbusular mycorrhizal fungi (abbreviated as AMF in this report). The AMF lives symbiotically with the plants - it provides the P to plants in exchange of CO2. It is well-known that plants have the tendency to grow their roots to the P-enriched soil areas. The main goal of this report is to investigate the interactive effect of AMF and P fertilizer on root allocation in P-limiting soil. Since the AMF improves the P-status and decrease the P-limitation of plants, we would expect that preferential allocation of roots towards the P-enriched soil be decreased.

Full Report [pdf]

Umfrage zu industriellen Dienstleistungen

Student: Al Maturo [al.maturo at]

Client: Karin Renninger (Univ. St. Gallen)


Abstract: Auswertung und Analyse einer Unipark-Umfrage zu Barrieren in Unternehmen und industriellen Dienstleistungen.

Full Report [pdf]

FS 2010

Scooter Toxicity

Student: Adrian Schmid [adschmid at]

Client: Loretta Müller (Uni Bern)


Abstract: Die Arbeit "Toxicity of Scooter Emissions" befasst sich mit dem Einfluss von Scooter- und Autoabgasen auf Zellen. Durch den Einsatz verschiedener Kraftstoffe, Filter und Fahrzeugmodelle werden unterschiedliche Abgase erzeugt. Die vorliegende Auswertung prüft den Einfl uss der physikalischen Eigenheiten und Zusammensetzungen der Abgase auf die Zellen.

Full Report [pdf]

Resulting Publication: Loretta L. Müller, Pierre Comte, Jan Czerwinski, Peter Gehr, Markus Kasper, Andreas C.R. Mayer, and Barbara M. Rothen-Rutishauser: Higher Toxic Potential Of 2-stroke Scooter Exhaust Emissions Compared To 4-stroke Scooter And Diesel Car Emissions, Am. J. Respir. Crit. Care Med. 181: A6826.

Calibration of two Soil Analysis Methods

Student: Alexios Kariagiannis [alexiosk at]

Client: Ubald Gasser (Kanton Zurich, Fachstelle Bodenschutz)


Abstract: This report includes the comparison of two chemical extraction methods in soil. The goal is to study if the cheaper method could efficiently replace the expensive one:
Cation exchange capacity and Exchangeable Cations are measured in soil samples by a procedure called "FAC". This procedure costs CHF 300 per sample. There also exists another method called "AAE-Extraction" (AAE) in which extraction of the nutrient elements is done. This procedure costs CHF 50 per sample. Consequently, there is a strong argument in favor of the AAE method. The purpose of this project is to analyze whether it is possible to substitute-predict FAC from AAE. These predictions concern the following chemical elements: Aluminum (Al), Calcium (Ca), Kalium (K), Magnesium (Mg) and also kakeff which shows the cationexchange capacity. Kakeff represents the quantity of cations bound to soil colloids, taking into consideration the strongly pH-value-dependent charge of the organic substances. For the prediction of kakeff all the elements will be used.

Full Report [pdf]

Earthquake Early Warning

Student: Michel Philipp [ at]

Client: Georgia Cua (ETH Zurich, Earthquake Statistics Group)


Abstract:In the field of earthquake early warning it is of fundamental interest to detect earthquakes immediately after occurance by analysing the seismic measurements of the ground motion. In Switzerland, this is investigated by the Swiss Seismological Service with the latest "Multiple Threshold Event Detection" approach, where few seconds of a seismic waveform sequence is aggregated into 15 parameters and inspected in order to filter out ground motion from non-earthquake sources and detect earthquakes immediately after occurance.
This report describes the model building process and the resulting error rates
of two statistical methods, the logistic regression and random forests, applied to the classification of ground motion events into earthquakes and non-eqartquakes. Both methods show that the most important parameters to discriminate between the two classes are the maximum and average from velocity, acceleration and displacement of the arrived waveform sequence. The model with the best performance is based on the random forests method and a 3 second waveform sequence.
As the model predicts probabilities for an earthquake, the resulting error
rates depend on a threshold that can be chosen individually. Setting this "cut-off"at 0.5, the probability for detecting an earthquake is 78.1% and the probability of filtering out non-earthquakes is 93.6%. To filter out 99% of  non-earthquakes, the "cut-off" must be increased to at least 0.75. The probability of detecting an earthquake then declines to 51.0%. Although these results are satisfying, there are several options to further improve the model building process.

Full Report [pdf]


Wichtiger Hinweis:
Diese Website wird in älteren Versionen von Netscape ohne graphische Elemente dargestellt. Die Funktionalität der Website ist aber trotzdem gewährleistet. Wenn Sie diese Website regelmässig benutzen, empfehlen wir Ihnen, auf Ihrem Computer einen aktuellen Browser zu installieren. Weitere Informationen finden Sie auf
folgender Seite.

Important Note:
The content in this site is accessible to any browser or Internet device, however, some graphics will display correctly only in the newer versions of Netscape. To get the most out of our site we suggest you upgrade to a newer browser.
More information

© 2016 Mathematics Department | Imprint | Disclaimer | 31 August 2011