APPENDIX
SOME RECOMMENDATIONS.
In general, an experiment passes through five phases in the following order:
(1) specification of the purpose of the experiment,
(2) planning of the experiment,
(3) conduct of the work,
(4) analysis of the data, and
(5) presentation of the experiment.
The following suggestions are offered with diffidence, since the writer's
experience both in statistics and in photogrammetry is very limited. It is however
hoped that these suggestions stimulate further discussions on the requirements in
experimental research in photogrammetry, and on the ways in which these requi-
rements can be realized most efficiently, in order to avoid such shortcomings as
have been disclosed in this paper.
A.1. SPECIFYING THE OBJECTIVES OF THE EXPERIMENT.
The objectives of the experiment should be specifically and clearly defined.
The main objectives are of course those of immediate interest to the expe
rimenter. An attempt should however be made to cover issues of more
general interest whenever this is practicable.
A.2. PLANNING OF THE EXPERIMENT.
The purpose of planning is to ensure that the experiment shall realize its
objectives as economically as possible.
A.21. Choice of Variates.
It frequently happens that the experimenter has to choose from a number of
plausible criteria. Selection must be based on balancing the following requi-
rements in the variates which would be obtained by applying these criteria:
(1) The variates should be most relevant to the main objectives of the
experiment.
(2) Their interpretation should not be ambiguous.
(3) They should satisfy the assumptions underlying the proposed method
of analysis, or can be suitably transformed.
If this condition cannot be fulfilled, a different form of analysis should
be sought for before proceeding any farther.
A.22. Recording the Factors that may Influence the Variates.
All the factors that can conceivably influence the chosen variates should be
specified in detail. In general these factors may be classified as:
(1). Methods or treatments by means of which the variates are derived from
the observations (if the variates are not themselves the observed
quantities).
(2). Instrumental errors.
(3). Observational errors, including human influences.
(4). Physical conditions which can be controlled at pre-determined levels.
(5). Physical conditions which can be determined but cannot be controlled.
(6). Experimental conditions which can neither be controlled nor determined.
A.23. Design of the Experiment.
A simple factorial scheme of the experiment may well serve as a tentative
27