Conference: Methods for Longitudinal Data Analysis in the Social Sciences

Monday 8 September 2014, 10:00 – 18:00
Followed by drinks reception and buffet dinner

Conference venue: New Theatre, East Building, Houghton Street, LSE campus


*** Limited additional places are now available to attend this conference. Please email Ian Marshall if you would like to reserve a place. ***


These are unprecedented times for the social sciences in terms of the availability of high-quality longitudinal data. The richness of these data are enabling researchers to broaden their horizons and contemplate addressing research questions of increasing complexity, and develop models of increasing sophistication to answer these questions.  In this event, we bring together researchers from social statistics, biostatistics and economics to talk about some of the latest developments in this area.  The speakers will talk on a range of subjects, from innovative ways of collecting longitudinal data to dealing with its most difficult problems, from modelling growth and over-dispersion to estimating causal effects.

This event is sponsored by the LEMMA node of the ESRC National Centre of Research Methods, and is organised by Fiona Steele (LSE) and Paul Clarke (University of Essex).

Speakers

Paul Clarke (University of Essex)
Dynamic panel-data modelling with structural nested mean models

Marcel Das (CentERdata / Tilburg University)
Innovation in online longitudinal data collection for scientific research

Bianca De Stavola (London School of Hygiene and Tropical Medicine)
Mediation and life course epidemiology: Challenges and examples

Harvey Goldstein (University of Bristol)
Modelling repeated measures growth data by aligning significant growth events and modelling changes in within-individual variability over time

Geert Molenberghs (Katholieke Universiteit Leuven)
A flexible modelling framework for over-dispersed, hierarchical data of a joint nature

Anders Skrondal (Norwegian Institute of Public Health)
Protective estimation of panel models when data are not missing at random

Tom Snijders (University of Oxford, University of Groningen)
Longitudinal methods for using panel data of networks and behaviour to assess peer influence

Please view abstracts of the talks HERE

Advantages and disadvantages of secondary data collection nowadays

Advantages

1. The first advantage of using secondary data (SD) has always been the saving of time (Ghauri, 2005). Not enough with this, in the so called Internet Era, this fact is more than evident. In the past, secondary data collection used to require many hours of tracking on the long libraries corridors. New technology has revolutionized this world. The process has been simplified. Precise information may be obtained via search engines. All worth library has digitized its collection so that students and researchers may perform more advance searches.

2. Accessibility. In the past, SD was often confined to libraries or particular institutions. Top of that, not always general public gained access. Internet has especially been revolutionary in this sense. Having a internet connection is frequently the only requirement to access. A simple click is sometimes more than enough to obtain vast amount of information. The problem, nevertheless, is now being able to see whether the data is valid.

3. Strongly connected to the previous advantages is the saving of money (Ghauri, 2005). In general, it is much less expensive than other ways of collecting data. One may analyzed larger data sets like those collected by government surveys with no additional cost.

4. Feasibility of both longitudinal and international comparative studies. Continuous or regular surveys such as government censuses or official registers are especially good for such research purposes. The fact of being performed on a regular or continuous basis allow researchers to analyze the evolution of, to give an example, per capita income in Poland from 2000 to 2012. Something similar occurs when comparing different countries. Although important difference between countries may exist, the truth is that censuses and other government studies tend to unify criteria all over the world or, at least, within certain geographical areas, such as European Union, or among certain international organizations members, such as OECD. Another example are the studies carry out by international networks that aims to collect information world-widely following the same criteria. The World Values Survey is a good example. It is a source of empirical data on attitudes covering a majority of the world´s population (nearly 90%) It is carried out by a worldwide network of social scientist who, since 1981, have conducted representative national surveys in almost 100 countries. Aiming such data for international or longitudinal studies via primary data collection is truly difficult and often miss the rigor that diverse social contexts comparisons require.

5. Generating new insights from previous analyses (Fàbregues, 2013). Reanalyzing data can also lead to unexpected new discoveries. Returning to the previous example, the World Values Survey Association usually publish the so called World Values Survey Books. They are a collection of publications based on data from the World Values Surveys. Since the database used may be accessible for outsider, you can analyze the data and come up with new relevant conclusions or simply verify and confirm previous results.

Disadvantages

1. Inappropriateness of the data. Data collected by oneself (primary data) is collected with a concrete idea in mind. Usually to answer a research question or just meet certain objectives.  In this sense, secondary data sources may provide you with vast amount of information, but quantity is not synonymous of appropriateness. This is simply because it has been collected to answer a different research question or objectives. (Denscombe, 2007). The inappropriateness may be, for instance, because of the data was collected many years ago, the information refers to a entire country when one aims to study a specific region, or the opposite, one aims to study an entire country but the information is given in a region wide. There are two possible ways to be taken when SD is not appropriate: 1) answering your research question partially with the subsequent lack of validity; 2) you need to find an alternative technique of data collection, such as survey or interviews.

2. Lack of control over data quality (Saunders, 2009). Government and other official institutions are often a guarantee of quality data, but it is not always the case. For this reason, quality issues must be verify as outlined in this post.

Any other advantage or disadvantage up? Would you like to add something else?

Related articles

Reference list

Denscombe, M. (2010). The good research guide: for small-scale social research projects. Open University Press.

Fàbregues, Sergi (@sfabreguesf). “@socioloxia  Perform an alternative analysis with the aim of generating new insights from previous analyses” 10th of December 2013, 11:33 AM.

Ghauri, P. N. (2005). Research methods in business studies: A practical guide. Pearson Education.

Saunders, M. N., Saunders, M., Lewis, P., & Thornhill, A. (2011). Research Methods For Business Students, 5/e. Pearson Education India.