Abdul Mckenzie posted an update 11 months, 4 weeks ago
Influences on clutch size, brood size and numbers of males in a clutch were identified using log-linear analyses, which are appropriate for small count data. Influences on sex ratios, the prevalence of all-female broods, and developmental mortalities were explored using logistic analyses, which are appropriate for proportional data. We report the percentage of deviance explained (%Dev) as an approximate analogue of r2 for log-linear and logistic models. We used backward stepwise procedures and aggregation of factor levels to obtain the parsimonious ��minimal Pfizer Licensed Compound Library nmr adequate model�� by model simplification (e.g. Crawley 1993; Wilson & Hardy 2002). We used the variance ratio, R, as a descriptive statistic to quantify the variance of brood sex ratios and developmental mortality across broods: R?=?1 indicates a binomial distribution, while R??1 indicate under- and over-dispersion, respectively (Krackow, Meelis & Hardy 2002). The Meelis test statistic, U, was used to assess the significance of any deviations from binomiality (Krackow, Meelis & Hardy 2002). Overall, we collected 2758 parasitized scales; 494 of these scales were parasitized by M.?luteolus. The remaining scales were parasitized by other Metaphycus species or by species belonging to other genera (Kapranas et?al. 2007). A few scales yielded both M.?luteolus and another parasitoid species, but this occurred rarely (n?=?7). In some cases, scale size and/or parasitoid mortality factors could not be assessed, but each part of the analysis was conducted using all suitable data available. There was a total data set of 395 clutches for assessing mortality, of which 371 were gregarious (clutch size?>?1). The number of eggs laid in each host (=?clutch size) ranged between one and nine, although 95% of clutches contained 1�C6 eggs, and clutch size increased with host size (log-linear analysis corrected for underdispersion; F1,403?=?289��72, P?<?0��001, %Dev?=?41��9, Fig.?2). For clutches of one egg (solitary clutches) and without developmental mortality (n?=?24), the probability of the offspring being a male (sex ratio) decreased as host size increased (simple logistic regression; G1?=?11��20, P?<?0��001, %Dev?=?51��8, Fig.?3). For clutches of more than one egg (gregarious clutches) and with no developmental mortality (n?=?127), the proportion of offspring that were male (sex ratio) decreased significantly with clutch size (logistic regression corrected for underdispersion: F1,125?=?18��87, P?<?0��001, %Dev?=?13��0), whereas the number of males laid in each clutch increased (log-linear regression corrected for underdispersion: F1,125?=?12��25, P?<?0��001, %Dev?=?9��0). These analyses were carried out with the caveat that subsets of broods lacking developmental mortality may not be representative of the overall primary sex ratio (Fiala 1980). Thirty per cent (428/1449) of M.?luteolus offspring developing in gregarious clutches died before adulthood.