ANIMAL feeding trial experiments are performed to investigate the comparative effects on growth of different diets. The experiments are carried out over several generations of a large number of animals. Each experiment is, in general, arranged in a different statistical design, which is usually decided by the number of diets to be tested; but other effects such as litter and room position have to be considered. Thus, each experiment requires a different type of analysis.
Measurements such as weight, body length and food intake, are made on the animals at regular intervals; it is easy to see that, with several experiments running concurrently, a large amount of data is collected at the same time. This puts the analysis and data handling beyond the range of desk machines and human beings and into the realm of computers. In these trials, the observations for one animal in one experiment in one week are punched on to one card. A program has been written which assembles and analyses this weekly data as it is provided and stores it on magnetic tape. In so doing, the information is made easily accessible for any future reference.
The program calls upon three main subroutines: BASIC, OBSER and QUEST. To each animal in the experiment is attached a number and a set of parameters defining its position in the experiment and the diet it is to receive. This information is punched on to one card, and the set of cards for all animals constitutes the Basic Data. This is introduced by headings which reveal general information about the experiment and the first card of the Basic Data deck contains the word 'Basic'. When the main program reads this word the subroutine BASIC is called and it is this routine which reads and performs checks on the Basic Data. When this has been satisfactorily completed the information is written to magnetic tape.
In the laboratories the weighing machine is connected automatically to an IBM 026 punch, which produces the card containing the observations. Again, there is one card for each animal in the experiment. The week number is also punched on this card. The word 'Observations' introduces a deck of such cards and any general information, such as experiment and generation numbers, as well as information about the death of animals, is punched on cards which immediately follow this card. The subroutine OBSER compares each observation against previous measurements and if any large discrepancies occur a message is output. A check is made that all observation have been presented and then the information is written to tape and an analysis of variance performed. The results of the analysis appear on a print-out.
The experimentalist is able to retrieve information about animals, litters or treatments etc. at any time during the experiment or after its completion, by presenting just one card to the program. Growth curves may be similarly obtained. The word 'Questions' introduces such cards and the routine which interprets these cards is called subroutine QUEST.
An illustration of a possible data deck will now be given. (One line represents one card.)
BASIC DATA EXPERIMENT 28 GENERATION 0 STARTING AGE 4 MALES 32 FEMALES 32 BIRTHWEEKS 132 TO 132 MEASUREMENTS WEIGHT LENGTH TAIL ANALYSIS 3 ITEMS 10 POSITIONS WEEK 2 RAT 4 WEIGHT 6 LENGTH 8 TAIL 10 DOSES 2 HEIGHTS 4 COLUMNS 4 STARTWEEK 136 ANIMAL RAT NAME FISH TOXICITY TEST INTERVALS WEIGHT 1 LENGTH 1 TAIL 1 1 2975 524 2 3 3 2 M Basic data for 2 2994 527 2 3 3 2 M Experiment 28 . . . Generation 0 64 5064 538 19 1 2 R F OBSERVATIONS EXPERIMENT 28 GENERATION 0 WEEK 136 RAT 2975 WEIGHT 75.1 LENGTH 21.6 TAIL 6.4 WEEK 137 RAT 2975 WEIGHT 79.2 LENGTH 22.1 TAIL 6.5 Observations WEEK 138 RAT 3013 WEIGHT 70.4 LENGTH 20.9 TAIL 6.2 QUESTIONS EXPERIMENT 27 GENERATION 1 EXTRACT VARIABLE WEIGHT FOR ANIMAL 5064 FROM WEEK 71 TO WEEK 104 FINISH
The word FINISH terminates the data deck. It should be pointed out that each of the three sections above can be presented separately but observations cannot be presented before the Basic Data for a given experiment and generation is on tape, and questions cannot be asked until the required information has been presented as observations.
One looks forward to the introduction of typewriters at the Atlas Laboratory. If one could be situated in the biological laboratories and connected by land-line to the machine it would be possible for the experimentalist to have observations automatically checked as he does his measuring rounds and to receive the results of analyses in a matter of minutes.