STATISTICAL EDUCATION AT THE UNIVERSITY OF NATURAL

RESOURCES AND LIFE SCIENCES IN VIENNA

Karl Moder

Institute for Statistics and Computing, University of Natural Resources and Life Sciences

Gregor Mendelstrasse 33, A-1180 Vienna, Austria

Keywords:

Assessing student learning, Online programme, Large class size.

Abstract:

Introductory statistics courses at the University of Natural Resources and Life Sciences in Vienna suffer from a

very high number of participants. Correcting written tests is almost impossible because of a lack of personnel

resources. As an alternative an online programme has been developed. It considerably reduces the time

required by instructors and it seems to intensify the cooperation among students.

1 INTRODUCTION

At the University of Natural Resources and Life Sci-

ences in Vienna the number of newly enrolled stu-

dents increased from 696 to 2584 within the last 9

years. About 20% out of these 2584 students are non

Austrian citizens. The ratio of male to female students

changed within this time period from 1.11 to 1.01.

During the winter semester 2010/2011 and depending

on their branch of study 1534 out of 2584 students

had to attend an introductory statistics course which

was supervised by the scientiﬁc staff of the statistical

institute.

The following table shows the branches of study

and the corresponding number of students.

Table 1: Branches of study and corresponding number of

attending students in 2010/2011.

number of

branch of study students

Agricultural Sciences 297

Environment and Bio-Resources

Management

461

Food Science and Biotechnology 374

Landscape Architecture and

Landscape Planning

343

Viticulture, Enology and Wine

Economics

1534

At the moment the scientiﬁc staff of the statistical

institute consists of 3 people. Because of this small

number of statisticians, students from the ﬁrst and last

branch of study shown in table 1 are taught together.

Since there are students from higher semesters who

failed this course in previous years, the number of

participants in the courses increased, e.g. in Environ-

ment and Bio-Resources Management to more than

500. Because of the low personal equipment and the

high number of participants of introductory statistics

courses an online-program was developed to reduce

the burden of statistical education for the staff of the

statistic institute.

2 OUTLINE OF STATISTIC

COURSE

The following description of the statistics course is

based on lectures of the author for students of ”Land-

scape Architecture and Landscape Planning”. Stu-

dents attend this course in their ﬁrst semester of stud-

ies. In 2010 a total of 351 students (60% female, 40%

male) registered to the course. About 25% of all par-

ticipants abandon the course within the ﬁrst 5 weeks

because they leave university. The drop out rate for

the remaining students is about 7% – 9% and rather

stable over the last 10 years.

Roughly 50% to 60% of students attend the lecture

until the end of the term. There is no obligation to at-

tend the lecture; this ﬂexibility is permitted because

about 50% of students have a job beside their stud-

ies. The lecture takes two hours per week during the

whole semester. Its contents are based partly on books

like (Essl, 1987) or (Rasch et al., 2007). A script can

Moder K. (2011).

STATISTICAL EDUCATION AT THE UNIVERSITY OF NATURAL RESOURCES AND LIFE SCIENCES IN VIENNA.

In Proceedings of the Second International Conference on Innovative Developments in ICT, pages 77-81

DOI: 10.5220/0004472000770081

 SciTePress

be downloaded from the ELearn platform of the uni-

versity or can be bought at the institute. In addition to

this script there are several examples, pictures and an-

imations which can be downloaded, too. All lectures

are recorded via lecturnity software and can either be

be watched on the Internet or downloaded.

Until 10 years ago students had to pass 2 written

tests and a ﬁnal oral exam. Since this was very time

consuming, an online programme was developed for

examination purposes.

This means, at the moment statistical education

consists of three parts:

• Students have to chose 10 online examples and

complete the task presented.

• They have to complete 3 different tasks in SAS.

• Finally they have to pass a written test, consisting

of 2 parts. Part 1 concerns statistical basics. In

part 2, 5 – 6 data sets are provided and students

have to determine which method to use to solve a

certain statistical problem.

As mentioned, this is for students of ”Landscape

Architecture and Landscape Planning”. For other

branches of study the number of online examples is

12, and 4 tasks have to be completed with SPSS or

R. This paper describes a programme which creates

online examples that students can use to practice, to

accomplish the required tasks and receive feedback

until they are ready to take a ﬁnal test.

2.1 The Online Programme

One prerequisite for the use of this programme is a

computer and an internet connection. Surveys among

students of Landscape Architecture and Landscape

Planning in the last 3 years showed that all of them

had the necessary access. Those who think they might

have problems using a computer can attend a basic IT-

course; in the past, hardly anyone did so. Addition-

ally, a detailed description of SAS, SPSS and R nec-

essary for the exercises, that students have to do by

using those programmes can be downloaded from the

ELearn-platform of the university. All of these pro-

grammes and introductions are located on servers and

can be accessed from home (and are also available for

readers of this article).

To attend the statistic course, students have to en-

roll by name, branch of studies and student ID. These

data are stored in a MySQL database. After enrol-

ment, students are allowed to use the online pro-

gramme. At the moment, the programme offers the

following 14 possibilities.

By modifying a list of indexes in the PHP pro-

gramme the order as well as the number of tasks can

Table 2: Methods offered by the online programme.

descriptive methods

stem & leaf diagram

boxplot

parametric tasks

conﬁdence intervals

sample size determination

one sample t-test

two sample t-test

one way ANOVA

two way ANOVA with interaction

paired t-test

block analysis

regression and correlation

non parametric tasks

contingency tables

Kolmogorov-Smirnov-test

Kruskal-Wallis-test

be changed easily. In a second list the number of

points which students can obtain for each example is

determined. Students may only sit the written exam

if the reach at least 50% of the maximum number of

points obtainable for every example they have dealt

with. 80 to 88 percent of participants archieve more

then 90% of the maximum number of points for each

example. To use the online material students have to

identify the programme and select the exercise they

want to deal with (ﬁgure 1).

Figure 1: Identiﬁcation and task selection frame.

After a task has been selected, input parameters

are listed. If ”Create a dataset” is then chosen and as

soon as input is conﬁrmed, each student gets his/here

own data set of his own (ﬁgure 2).

These data sets are created rather ﬂexible. For ex-

ample, if one selects t-test, 10 different texts are avail-

able at the moment (in principle, there is no limit to

the number of texts). For the creation of data sets, the

following parameters have to be set to obtain plausi-

ble ”observations”:

1. minimum and maximum number of observations

for the ﬁrst (e.g. 7 – 11) as well as for the second

INNOV 2011 - Second International Conference on Innovative Developments in ICT

Figure 2: An individual data set for a certain student and

exercise.

sample (e.g. 7 – 11),

2. decimal points for observations (if this value is

set as ”1” all digits following the ﬁrst digit after

the decimal point are set as”0”, e.g. 27.1 means

27.10000),

3. a distribution for the generation of observations

(normal distribution or uniform distribution are

available at the moment),

4. a range for the observations according to each text

(for the example above that would be 30 to 60).

These data are stored in an external deﬁnition ﬁle and

can be changed with the help of any text editor. Four

different α values are chosen randomly (0.1, 0.05,

0.01, 0.001) by the programme. For a t-test, a one-

sided or two-sided test situation is chosen at random.

The programme is adjusted in such a way, that about

20% of datasets lead to a signiﬁcant result (but this is

different for different examples and can be modiﬁed

easily).

Within a certain time limit (7 – 14 days) students

have to accomplish their task and enter their results

(ﬁgure 3). If one uses a registration number from 1 to

99 there is no time limit for any task, but in that case

results are not considered for grading purposes. Most

of the frames require the input of calculated results as

well as some basic theoretical answers.

Figure 3: Input frame for the results.

Once a student has entered his answers he/she can

check if they are correct by clicking the ”Check re-

sults” button. By doing so, students obtain a table

showing correct and incorrect answers (ﬁgure 4). A

green ﬁeld beside a value indicates a correct result, a

red one an error. As generated data are exact in the

sense that all digits behind the decimal point which

are not shown are 0 (with some limitation to the repre-

sentation by the computer (McCullough, 1998)) there

are only very few problems with rounding errors if

students use a calculator or SAS, SPSS or R, because

the online programme tolerates a deviation from the

exact value of up to 1% (this value is deﬁned in the

PHP programme and can be altered without any prob-

lem). The tolerance value should not be too large;

for example in a one way analysis of variance stu-

dents compute the sum of square value for total (SS

)

and for the factor (SS

). The sum of square value for

the error term (SS

) is calculated as the difference be-

tween these two values. Even if the value for SS

wrong by less than 1% this can lead to a wrong SS

In any next step the deviation from the exact value is

possibly bigger than 1% in relation to the size of this

value. So students get confused because SS

and SS

seem to be correct but the difference (SS

) is not.

If students are satisﬁed with their results, they can

save them. Otherwise they can go back and correct

their input.

Figure 4: Check validity of entered data for a speciﬁc ex-

ample.

At each point in time students can get information

about the total score they have so far achieved until

the last input (ﬁgure 5).

Within a time window of about 7 to 14 days stu-

dents have to complete a certain task. Figure 6 shows

the percentage share of ﬁnished tasks over time for 3

different online examples (block analysis, regression

analysis and contingency tables).

Whenever block design is used, the input has to

be ﬁnished within 10 days, for regression analysis and

contingency tables there are 12 days available for in-

STATISTICAL EDUCATION AT THE UNIVERSITY OF NATURAL RESOURCES AND LIFE SCIENCES IN

VIENNA

Figure 5: Overall score up to the last input.

Figure 6: Percentage of ﬁnished examples over time (day of

input).

put. A regression has to be worked out with SAS.

Maybe this is one reason, why curves in ﬁgure 6 dif-

fer that much.

In a ﬁnal written test students have to identify an

appropriate statistical method to analyse a speciﬁc

data set. With the help of the programme they can

enhance their skills to accomplish this (ﬁgure 7).

Figure 7: Test preparation programme.

One out of 8 non descriptive statistical methods

(table 2) is randomly selected by the programme and

for this method one out of about 600 data sets is cho-

sen randomly. In some situations (t-test - analysis of

variance, paired t-test - block analysis) multiple an-

swers are possible. Each answer is scored and all

scores are summed up until the last task.

2.2 Realization of the Programme

Data sets and corresponding solutions are created by

means of a FORTRAN-programme (Vaught, 2006).

These data are integrated into a MySQL-table. Cor-

responding PHP programmes produce several online

masks to present tasks, enter solutions, check results

and practice for the written test. The programme is

rather modular and can easily be extended to other

statistical methods. The language is German, as Aus-

tria is a German speaking country. But the language

can be changed easily, because text is stored in a

database and is independent from the generated data

(There are some restrictions to the range of created

data, to make sure that they are plausible in the con-

text provided). Students are invited to invent suitable

texts for online examples. For each text that is ac-

cepted (a maximum of 4) they are credited 1 point for

that part of the ﬁnal test where they have to select an

appropriate analysis method (the maximum score for

that part is 20 points and a minimum of 10 points is

required to pass). About 30% of the students who ﬁn-

ished the course submitted some text. Most of them

sent 4 texts and 2.1 texts were accepted on average.

3 CONCLUSIONS

Several tools to enhance statistical skills (mostly Java

applets) (e.g. (Dariu, 2006)) and a lot of supplemental

material for statistics courses (e.g. (Pezzullo, 2011))

can be found on the Internet. Multiple choice tests

can be developed with the help of ELearn platforms.

Details about multiple choice randomization can be

found in (I. McLeod, 2003). Other forms of on-

line teaching are demonstrated in (Tudor, 2006) and a

comparison between online teaching and class teach-

ing can be found in (Dutton and Dutton, 2005). But

to my knowledge, no course comparable to the one

presented here is available.

Experience gained over the last ten years leads to

the following impressions:

• Although only 50% of the maximum score need

to be achieved to be permitted to sit the ﬁnal writ-

ten test, between 80 and 88 percent of participants

reach more then 90% of the maximum score.

• Students try to improve their outcomes.

INNOV 2011 - Second International Conference on Innovative Developments in ICT

• More then 4% of the students improve their re-

sults within the input time window although they

already gained a sufﬁcient number of points with

their ﬁrst input. Three of them did this 3 times.

How often results are entered within a day cannot

be evaluated.

• The programme is widely used to practice the de-

termination of appropriate statistical methods for

the analysis of a given dataset.

• Drop out rates among students as well as their

grades are very similar for the online programme

and the old version of the course (about 7% to

10% of those who continue their studies at the uni-

versity).

• Correspondence via email and within the scope of

an online forum shows that the programme en-

courages students to cooperate. There are more

than 300 entries in this forum, where students on

the one hand ask for help while others post an-

swers to their questions.

The online programme started in 2003 for one branch

of studies and is now used by various lecturers in 4

different branches of studies. In 2010 the course was

evaluated as one of the ten best lectures of the Univer-

sity of Natural Resources and Applied Life Sciences.

No commercial software is needed to run the pro-

gramme. All online masks and database connections

are based on PHP. For database management MySQL

is used. To run the programme on the World Wide

Web, access to a webserver is necessary. Data and

solutions for online examples are generated by means

of a Fortran programme, but can be done in any pro-

gramming language. Since the experiences with this

kind of online teaching has been very positive, it will

be continued. The source code for the online pro-

gramme is available from the author. Since other

disciplines beside statistics show interest in the pro-

gramme, it is planed to reprogramme the source in

Java in such a way that online masks can be generated

without any knowledge of a programming language.

There are 2 variations of the programme depend-

ing on the branch of study. If one enters a reg-

istration number between 1 and 99 the online pro-

gramme can be used by everyone without any re-

striction (http://statlap.boku.ac.at/index.php). As far

as the theoretical part of the ﬁnal test is concerned a

multiple choice module containing theoretical ques-

tions is planned.

Three datasets of the online programme have to

be analyzed by means of a statistical package (SAS,

R, SPSS). The results with suitable explanation have

to be uploaded to an ELearn platform and results ob-

tained with the help of these packages have to be en-

tered into the online programme, too.

The lecture is accompanied by an ELearn course

where students get some help regarding all aspects of

the programme and the lecture. An online forum ex-

ists, which is widely used by participants to commu-

nicate with each other and with lecturers.

To sum up it seems that the program presented

embedded in several introductory courses is very suc-

cessful. The burden for the statistical staff is bearable

and the quality of education seems to be comparable

to former practice.

REFERENCES

Dariu, P. (2006). Environments to experiment.

http://lstat.kuleuven.be/env2exp/.

Dutton, J. and Dutton, M. (2005). Characteristics and per-

formance of students in an online section of business

statistics. Journal of Statistics Education, 13(3).

Essl, A. (1987). Statistische Methoden in der Tierproduk-

tion.

Osterreichischer Argrarverlag, Wien.

I. McLeod, Y. Zhang, H. Y. (2003). Multiple-choice ran-

domization. Journal of Statistics Education, 11(1).

McCullough, B. D. (1998). Assessing the reliability of sta-

tistical software: Part i. The American Statistician,

52(4):358–366.

Pezzullo, J. C. (2011). Statpages.net. http://statpages.org/.

Rasch, D., Verdooren, L. R., and Gowers, J. I. (2007).

The Design and Analysis of Experiments and Surveys.

Oldenborg Wissenschaftsverlag Gmbh, Wien.

Tudor, G. E. (2006). Teaching introductory statistics online

satisfying the students. Journal of Statistics Educa-

tion, 14(3).

Vaught, A. (2006). G95. http://www.g95.org/.

STATISTICAL EDUCATION AT THE UNIVERSITY OF NATURAL RESOURCES AND LIFE SCIENCES IN

VIENNA