and Summary In ethological studies it is common practice to obtain multiple observations on each individual in a sample and to Pool these observations into one data set for statistical analysis. In this paper we argue that this procedure reflects a fundamental error in the logic underlying random sampling since it implicitly assumes that the purpose of data gathering in ethology is to obtain large “samples of behaviour” rather than samples of behaviour from a large number of individuals. That is, it is assumed that the reliability of estimates of population parameters can be increased by obtaining additional observations on individuals already in the sample rather than by increasing the number of individuals observed. Using a Monte Carlo simulation we show that when such Pooled data sets are analysed statistically, the probability of rejecting a true null hypothesis is almost always substantially greater than the stated alpha level. Zusammenfassung In ethologischen Untersuchungen ist es üblich, von jedem Individuum einer Stichprobe mehrere Meßwerte zu gewinnen und für die statistische Analyse die Werte aller Individuen zu einer einzigen Stichprobe zu vereinigen („pooling”). Diesem Verfahren liegt, wie in der vorliegenden Arbeit gezeigt wird, ein fundamentaler Irrtum bezüglich der zufälligen Stichprobenauswahl zugrunde, da implizit davon ausgegangen wird, man müsse große Stichproben der Verhaltensweisen anstatt Stichproben des Verhaltens möglichst vieler Individuen sammeln. Damit wird angenommen, daß Populationsparameter zuverlässiger geschätzt werden können, wenn man mehr Beobachtungen am bereits in der Stichprobe vertretenen Individuen macht als wenn man zusätzliche Individuen beobachtet. Mit einer Monte–Carlo–Simulation wird gezeigt, daß die statistische Analyse solcher zusammengesetzter („pooled”) Stichproben fast immer eine wesentlich großere als die angegebene Irrtumswahrscheinlichkeit für („type I error”) ergibt.
In this paper aspects of the temporal and sequential patterning of pecks in the domestic chick are examined. To this end experiments were done in which young chicks were allowed to peck at pairs of coloured stimuli - the time of each peck being recorded automatically by a small computer. A particularly striking feature of the chicks' behaviour was the tendency of the pecks to occur in rapid bursts. Such bouts have been described (and quantified) for behaviours in a number of species, but the bout itself has proven to be a particularly difficult unit to define empirically. Accordingly a model is proposed which describes the order independent features of the intervals between pecks, and also objectively defines a bout criterion interval. This model assumes that not only do pecks tend to cluster into bouts but that the bouts themselves occur in clusters. Consequently two types of between bout intervals will be generated: "not pecking" intervals, whose mean length is long, and which separate bout clusters; and intervals of medium average length, separating bouts within a cluster. It is assumed that these between bout intervals are generated by two random (Poisson) processes of sharply differing rate constants. To test this model an Algol procedure was written which describes the survivorship curve of intervals between pecks in terms of components that can be adequately characterized as negative exponential. It is found that with large sample sizes three distributions are consistently delineated; the "not pecking" intervals are characterized by an exponential with a rate constant of order 10-3, the intervals within a cluster by one of 10-2 and the within bout intervals by one of 10-1 (if the latter generating process is assumed to have a dead time). In addition, a stable bout criterion ranging from 1.9 to 2.3 secs is defined. It is also argued that degeneracies in this model can occur when sample sizes are small - in particular that the intervals of medium length are either not distinguished from those belonging to the outer state of long intervals, or fail to be completely delineated from the two other components. Part II of this paper examines additional features of the pecking behaviour within the framework of this model. The question of how a colour is preferred is explored and it is found that preferred colours in these experiments receive from 1.6 to 3.2 times as many bouts, which are on average 1.1 to 1.3 times as long as those directed at the less preferred colour. Some information on sequencing of bouts is presented in which it is shown that there is a probability of about .60 of continuing to "bout" at the same position when the two stimuli are the same colour. This finding is combined with a measure of overall colour preference to allow fairly consistent predictions of the transition probabilities between bouts when the stimuli differ in colour. Upon examining the intervals within a bout it is found that although they peak strongly at .3 secs they can range between .1 and 2.3 secs. Through indirect evidence it is argued that the number of pecks within a bout also varies substantially. However, it is found that the distribution of intervals within a bout remains nearly invariant regardless of what colours the chicks are pecking. It is suggested that this invariance provides some objective verification for the concept of a "bout unit". However, upon closer examination, differences in the distribution of within bout intervals are found. These differences are interpreted as resulting from the rather surprising occurrence of mixed bouts, that is bouts during which pecks were elicited by both stimuli. Intervals within such mixed bouts were on average longer than those occurring in bouts where all pecks were directed at the same stimulus and consequently effected shifts in the mean and increased the spread in some of the within bout interval distributions. It is tentatively suggested that such mixed bouts are elicited when the chicks attend to cues upon which the stimuli cannot be distinguished. An important assumption of the model is that the interval distributions arising from the three generating processes overlap and that while a stable and obj ective bout criterion can be defined it does not provide a neat means of classifying intervals into either within or between bout intervals. Some intervals less than the bout criterion are assumed to belong to the processes generating between bout intervals. For this reason the total number of bouts defined by the criterion interval will underestimate the true number of bouts occurring - some bouts will be merged together when one of these very short between bout intervals occurs. The occurrence of mixed bouts, along with the substantial proportion of undetectable between bout intervals is shown to hinder extensive examination of the properties of the bouts defined by the criterion interval; it is argued that considerable ambiguity exists in the bouts so defined. The origin of this ambiguity is discussed within the framework of current studies in which bouts of behaviour, defined by a criterion interval, are used as a basic analytical unit.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.