Numerical Analysis Skills for Historians (NASH)

Image

http://www.twainquotes.com/Statistics.html

 

Dr Paul Atkinson of the University of Huddersfield Centre for the History of Public Health and Medicine announces the first two workshops in his AHRC funded (award AH/L011395/1) research support project Numerical Analysis Skills for Historians (NASH). If you have already expressed an interest in the series of events or you think you could benefit from additional training in quantitative methods read on.

 

I’m now in a position to give some details of our seminars. These will take two forms, one on the interpretation of data and the other on relationships between data series. At present we have arranged two seminars of the first form. In the next few weeks I will circulate details of the slightly more advanced seminar on relationships between data series (including regression analysis). I hope to provide details, too, of one more ‘interpretation of data’ seminar.

 

The initial seminars contain about four hours’ learning, running from mid-morning to mid-afternoon with a break for lunch. The style will be hands-on, using worked examples to understand how different concepts and methods work. A note on the subjects covered is attached.

 

London seminar

 

At University College London on Wednesday 4 June

 

Presenter: Dr Andrew Hinde, University of Southampton. Dr Hinde is Head of Southampton’s Division of Social Statistics & Demography. His extensive publications include frequently used texts on demographic methods and on the population history of England.

 

Edinburgh seminar

 

At the University of Edinburgh on Monday 23 June

 

Presenter: Professor John MacInnes, University of Edinburgh. Professor MacInnes is the ESRC Strategic Advisor on Quantitative Methods Training, part of the ESRC Quantitative Methods Initiative. He sits on the advisory board of ‘getstats’, the Royal Statistical Society’s Statistical Literacy Campaign, and has created a new statistical literacy course for students from all colleges and years at the University of Edinburgh.

 

There is no charge for attendance. I regret that we are not able to help with travel costs or provide refreshments. If you would like to book, please email me for a booking form at p.d.atkinson@hud.ac.uk.

 

I look forward to seeing you,

Paul Atkinson

 

Image

 

 

A brief summary of the curriculum for the events follows:

 

Content: interpretation of data seminars

 

Fundamentals

Variables, values and cases

Levels of measurement, categorical and continuous variables

Measurement and data capture as a social process, with substantial measurement error of various kinds. Validity and reliability.

Samples and populations.

Probability

 

Empirical description

Summary description of the distribution of the values for a variable, by numbers or visual methods.

Exploration of patterns of association between variables (theory, model, hypothesis generation).

Variable ‘time’ of special interest to historians: trends, rates, indices, risks/hazards.

Summary descriptions of association by numbers (correlation and regression coefficients; contingency tables) or visual methods (scatterplots, clustered bar or boxplot charts,)

Common problems with empirical description of sample data

Testing of a hypothesis, model or theory to see if it is consistent with the data collected.

Limits of testing: noisy data, common logical fallacies

 

Representativeness and generalisability

Historians almost always work with samples: how ‘representative’ these are:

– Random variation

– Selection effects

To the extent that a sample is not fully representative of the population it is drawn from, it is less safe to generalise from discussion of the sample to discussion of the population it is drawn from.

 

Inference from samples to populations

Random samples permit calculations of the probability that sample results obtained lie within a certain distance of the (unknown) result that would be obtained were it possible to measure the population.

‘Bayesian’ approaches

 

Data quality

Data always suffers from substantial amounts of measurement error.

Such error can be thought of as ‘noise’ in the data which obscures the ‘signal’.

When conclusions depend upon identifying patterns in the data that may turn out to be the product of the noise rather than a signal

 

Data presentation

Less is more, spurious accuracy and too much detail

Essential information: source, N, data availability, base for percentages, indices or rates.

Common errors.

 

 

Advertisements

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s