CLOSER was recently commissioned by the Economic and Social Research Council to carry out a review of available training for longitudinal researchers.
In this blog, the review’s author, Dr Jane Maddock, reflects on her findings, and her own experience as a developing longitudinal researcher.
“Tell me what a t-test is.”
Sweaty palms, pulse racing, flushed cheeks and out comes a mumbling, vague attempt at a response. This was my PhD interview. I was staring down the barrel of a PhD in epidemiology for the next three years armed only with the knowledge gained from my MSc introductory statistics module (read: working knowledge of means and standard deviations, and using the point and click options on SPSS. Stata, what’s that?). Fear crept in as the realisation that a passion for the topic would only get me so far. The steep learning curve that I had to navigate during my PhD was stomach-churning. I felt as though I was constantly being shot – Lamborghini-like – along rollercoaster tracks; travelling from 0 to 80 mph in less than two seconds. When I briefly stopped at the top of that doctorate rollercoaster, only then did I realise how far I had come.
I was one of the lucky ones. I had supportive supervisors and excellent office buddies. They pointed me in the direction of good short courses in regression and websites that provided step-by-step instructions for analysis, and they sent me their Stata code from which I could learn. I will never forget the interminable patience of Diane, a PhD student who was a couple of years ahead of me. She was with me when I first downloaded data from the UK Data Service, talked me through the data-cleaning process and gave me tips that only an experienced data user can offer. I lapped up as many short courses and seminars, and as much of Diane’s sage advice as I could. Then my PhD was completed.
My post-doc managers have been very supportive of my training pathway. However, my workload has increased, everyone is under pressure to publish and I am now responsible for students. I want to help them in the best way I can while making sure that I answer my own research questions using the best methods and most appropriate data, but time and money often get in the way. Handling the data within unfamiliar datasets is still a daunting task, but it is far less time-consuming. My training in regression has progressed to multi-level modelling and causal analysis, facilitated by online materials, my supervisor and a WhatsApp group of my PhD comrades. My ascent in training continues, but instead of being at the mercy of a Lamborghini, I am settled in a rather more sensible Renault.
Training in using rich but ‘messy’ data
The UK’s longitudinal studies are some of the world’s best and can help to answer pertinent questions about society and public health. However, in the 2017 ESRC Longitudinal Strategic Review, it was noted that despite existing initiatives, training capacity should be improved. With this in mind, where should you start if you have a question that these studies could answer?
This year the ESRC commissioned CLOSER to conduct a review of these training needs. We received 304 responses to our survey, and conducted focus groups and interviews with new supervisors and senior academics. These are the key findings:
- Successful analysis of longitudinal data involves being able to:
- handle/clean the data
- use basic quantitative statistical methods
- use advanced methods.
These do not always progress consecutively.
- There are clear gaps in training provision for initial data handling/cleaning and a lack of real-life, “messy” datasets to support training.
- Staff who support early career researchers should have better training themselves, for example through increased training provision for mid-career researchers.
- Access barriers to training must be removed. This could involve delivering training across multiple formats and increasing open-access training material.
CLOSER’s Learning Hub is addressing some of these concerns by providing free online material to users who are in the formative stages of understanding what a longitudinal is through to analysis. Importantly, the Learning Hub outlines real-life research case studies and has developed teaching datasets based on the 1958 National Child Development Study. In the coming months and years, CLOSER will be adding new methods to the Hub’s Analysis module, and building the bank of research case studies to span the breadth of topics covered by the UK’s longitudinal studies.
My hope is that through the findings in this review and with the development of the CLOSER Learning Hub my little Renault will accelerate, giving me the space required to support people at earlier stages – especially if they don’t have a Diane.
Read the summary of CLOSER’s review of quantitative analytical training needs for users of longitudinal studies report.
Dr Jane Maddock is a Postdoctoral Research Associate at CLOSER. Follow her at @JaneMaddockPhD.