Abstract:

In April/May 2005, a postal questionnaire was used to elicit views of all NARO scientists
and technicians about their current levels of skills in statistics and data management,
and to learn about their future needs in these areas. The questionnaire was forwarded
by the Monitoring, Evaluation and Policy Unit of NARO by e-mail to Directors of all NARO
institutes and Managers of ARDCs with a request to copy it to all their scientists and
technicians. Responses were received by 47% of the scientists and 50% of the
technicians. Although this does not constitute full coverage, we believe that the findings
reported below give a reasonable reflection of the true situation concerning NARO
researchers' views of their needs.
Most scientists (95%) were of the view that they would use a statistician's assistance in
their research work if one were available at their institute. The area where most
respondents (87%) expected to involve a statistician was in questionnaire surveys,
followed by livelihoods analysis (75%) and then by on-farm trials (69%) and on-station
trials (68%). This perhaps reflects lower confidence amongst respondents in survey
work compared to studies involving experiments. This finding is consistent with the
statistical topics on which respondents had received training. About two-thirds of
scientists and about 30% of technicians, had received training in biometrics or statistics,
but none mentioned having had training in survey methods and only eight mentioned
learning about sampling methods. Despite this, it is interesting to note that 63% of
scientists and 31% of technicians, claim to have "some" or "lots of" knowledge about
sampling methods. Perhaps this indicates a misconception that "sampling methods"
refers only to the ability to select a sample appropriately once the method is specified, or
thinking of sampling as representing just simple random sampling.
A surprising number of scientists (17%) said they had no knowledge of how to generate
tables of counts and percentages which are most relevant to the analysis of survey data.
This indicates that courses in biometrics and statistics, followed as part of a University
degree probably did not Include any survey components and maybe that the focus of
much NARO research has changed and broadened since the times when staff were
trained. Knowledge of other techniques was also rather weak amongst scientists with
many having no working knowledge, or only a little knowledge, of 2-way analysis of
variance (45%), of factorial designs (55%) and multiple regression (62%), despite these Statistical software packages mot' farriiiiarto-seieritists.wer_eaSPSS, Genstat and MStat
with about half the respondents having little to lots of knowledge in one of these
packages. With the recent purchase of Genstat by NARO, those with Genstat experience
will hopefully be able to share their skills with others in their Institutes. MStat is now
generally regarded as an obsolete package, so familiarity with its use will not be of much
benefit to NARO researchers and its further use should be discouraged. Whether or not
knowledge of SPSS is useful will depend on whether NARO purchases this package in the
future, or some other package such as STATA which has been incorporated into the
teaching programme at the Institute of Statistics and Applied Economics in Makerere
University.
Most respondents (74%) had been involved at some stage in planning a data recording
sheet. However very few persons (25%) had had training in how to organise data in a
spreadsheet so as to avoid errors during data entry. Procedures most commonly used
for checking data quality were scrutinising data collection sheets, manual checking of
paper records against computer versions and capturing errors during data analysis.
Standard operating procedures for data validation, and devoting sufficient time to this
activity seemed to be lacking. Only a few respondents (23%) were using Excel's
substantial facilities for data validation although the basics of this software were familiar
to most researchers.
The most important priority training area identified by (81% of) scientists was in
standard methods of statistics using appropriate software. About 40-50% also identified
research data management, integrating qualitative and quantitative approaches, and
interpreting results and reporting as one of their top three training priorities. For
technicians, training in both research data management (77%) and in standard methods
of statistics (76%) were in the top three priorities, followed by interpreting results and
reporting (59%). About two-fifths of technicians also listed design of on-station and on
farm trials as one of their top three priority training areas.
Of the 82% of respondents who expressed an opinion about areas of concern regarding
statistics and/or data management, 71% recommended that they be given training in
statistics and data management, while 22% recommended purchasing relevant software.
About a fifth of respondents also recommended employing a statistician/data manager at
their institute.
With respect to constraints to research in general, over 95% of scientists and over
90% of technicians identified software, training and support in statistics and data
being areas featuring commonly in experimental agricultural and biological research.