log linear analysis

log linear analysis

a technique of statistical analysis commonly used on CROSS-TABULATIONS of data. It transforms non-linear models into linear models by the use of logarithms. This is necessary because social data is often nominal or ordinal and therefore does not meet the assumptions needed by many statistical techniques (see CRITERIA AND LEVELS OF MEASUREMENT). It is a CAUSAL MODELLING device, involving setting up models to test against the data, successively adjusting the model till the best fit is found.
References in periodicals archive ?
Log linear analysis was carried out for each disorder with body mass index, family status, educational level, environment, living accompanied and employment set as confounders.
717 Table 3 Log linear analysis results Entropy Concentration [chi square] (%) (%) Major Depression Disorder 0.
We reduced the number of accommodation types to two in order to avoid violating the assumptions of log linear analysis and to maintain the power of the test.
Once these 8 responses are crossed by the variables of interest, the expected cell sizes all drop below 5, violating an assumption of log linear analysis (Howell 2002).
Log linear analysis (4) was used to determine whether the variables price, type and location of accommodation predicted levels of discrimination against the black auditors, either as main effects or in interaction with each other.
The aim of log linear analysis is to construct a model that accounts for the frequencies in the contingency tables with the least number of terms.
Once we know which orders of effects are significant, a second stage of log linear analysis investigates which specific effects within that order are significant.
2] test, logistic regression analysis and hierarchical log linear analysis using SPSS.
A hierarchical log linear analysis using a stepwise analysis of categorical variables [6], enabled the model illustrated in Figure 2 to be constructed.