Quantitative Longitudinal Data Analysis by Vernon Gayle Paul Lambert

Author:Vernon Gayle, Paul Lambert [Vernon Gayle, Paul Lambert]
Language: eng
Format: epub
Tags: Social Science, Research, Methodology, Reference
ISBN: 9781350188877
Google: K5EFEAAAQBAJ
Publisher: Bloomsbury Publishing
Published: 2020-12-10T04:35:12+00:00

Figure 27 Stata output: Coefficients F tests for the random effects panel model

For the variable wave, the ‘within-respondents’ effect (which is measured by the three dummy v ariables) F = 47.77. This is the same as the value of F in the fixed effects panel model reported in Figure 24. The F value in the fixed effects model has 3,21 degrees of freedom. The first number represents the k=3 explanatory variables in the fixed effects model. The total number of degrees of freedom for the model is 31 (i.e. the number of observations – 1). The model degrees of freedom equals 10; this is 3 (i.e. k explanatory variables) plus 7 (i.e. the number of individual respondents –1). There are 21 residual degrees of freedom (i.e. 31 total degrees of freedom – 10 model degrees of freedom).

In this tidy ‘toy’ example the models produce congenial results. As we will see later in genuine research examples using real panel data, the results from the models will not necessarily line up as neatly. The computation behind the estimation of the panel models will however follow these underlying principles.

Comparing different panel models: Example 2

In this section we use an extract of data from the BHPS in order to elaborate upon the techniques and the models that have been introduced earlier in the chapter. The extract of data covers the first 10 waves of the BHPS (1991–2001). The panel comprises men aged 25–35 in 1991 and working full-time. They are original members of the BHPS sample, by which we mean they are members of the main BHPS and have not entered the study as part of the territorial booster samples, or the ECHP. The outcome variable of interest is their usual net pay per month (£) in their current job (which has been adjusted for inflation). There are a number of explanatory variables relating to the respondent’s occupation, their level of education and their circumstances in childhood.

A compact codebook for the data is reported in Figure 28. The dataset contains the cross-wave person identifier pid, an indicator for the wave of the BHPS wave, a household identification number zhid and a linking variable zpno that helps to identify which person in the BHPS household the respondent is. The dataset also includes the respondent’s year of birth zdoby. The outcome variable zpaynu2 measures the respondent’s usual net pay per month (£) and has been adjusted for inflation.3 The number of hours normally worked per week in the respondent’s main job is measured by the variable zjbhrs. The Cambridge Scale Score (male) of the respondent’s present occupation is recorded by the variable zjbcssm (see Prandy, 1990). The Cambridge Scale Score (male) of the respondent’s father’s occupation when the respondent was aged 14 is recorded by the variable pacssm. A simple measure of the respondent’s level of education is recorded in the dummy indicator graduate. The respondent’s age at the time of the interview (age in years – 25) is recorded in the variable zregage.

Download

Copyright Disclaimer:
This site does not store any files on its server. We only index and link to content provided by other sites. Please contact the content providers to delete copyright contents if any and email us, we'll remove relevant links or contents immediately.