Trends and patterns: how to find them and can you believe them?

  • Published on

  • View

  • Download


Trends and patterns: how to find them and can you believe them?. Michael Wood My aim in this lecture is to. Give you an idea of the methods (especially statistical ones) that are used for analysing trends and making forecasts - PowerPoint PPT Presentation


<ul><li><p>Trends and patterns: how to find them and can you believe them?Michael Wood</p></li><li><p>My aim in this lecture is toGive you an idea of the methods (especially statistical ones) that are used for analysing trends and making forecastsDiscuss some of the problems and limitationsSuggest some things to check for</p></li><li><p>Data and statisticsData are facts, figures and informationYou can either collect data yourself (primary data), or more likely get them from a book or website (secondary data)Statistics are things that can be worked out from data like averages, percentages, correlations, etc. (The word statistics also refers to techniques for analysing data.)</p></li><li><p>Data and statistics about the pastMany sources e.g. see unit guide and (click on Data tab) the source will have a link to info about the meaning, derivation, limitations, etc. Read this!Always based on the past may be yesterday, but more likely last month or last year Often a reasonable understanding of the recent past is good enough, and likely to be all you can get.Can you get statistics about the future?</p></li><li><p>Are statistics always right?InflationRPI (retail price index) inflation in Dec 09 was 2.4% ( (consumer price index) inflation was 2.9%Both measure the average change in the prices of consumer goods and services purchased in the UK ( so can they both be right? Unemployment can be measured by counting benefit claimants or by a survey Do you think the answers will be the same?</p></li><li><p>Three surveys to check accuracy of the NRE telephone serviceAn NRE sponsored survey found that the answers were97% correctA Consumers Association survey used a sample of 60 calls, mainly about fares. The worst mistake was when one caller asking for the cheapest fare from London to Manchester was told 162 instead of the cheaper 52 fare which was available via Sheffield and Chesterfield. The percentage correct was 32%A reporter rang four times and each time asked for the cheapest route from London to Manchester. The proportion of the four answers which were correct was25%(Source: Breakfast programme, BBC1 TV, April 30 2002.)</p></li><li><p>Things to check with data and statisticsThe sample size and how selected.A random selection process usually best (e.g. Iraq death rate survey method, not adultery surveys in mags)Beware possible bias e.g. silent evidence Taleb (2008)How the statistic is defined (CPI vs RPI)If possible see if you can find alternative sourcesRemember that errors are almost inevitable try to get an idea of how big the errors are likely to beRemember that there are always chance fluctuations check statistical significance (see )</p></li><li><p>Methods of forecastingTime series look at the pattern over time of the thing you are trying to forecastCausal modelling take account of other variablesSimulations (e.g. of economy or global temperatures) like a very detailed causal modelExpert judgmentBest to ask several experts independently Delphi methodClairvoyance and time travelA real business opportunity!</p></li><li><p>Time series analysisPlastic rulersWhy plastic?Regression, moving averages, seasonal factors See most books on business statisticsMany more advanced methods!All assume that patterns in the past will continueBe careful if this is not likely!</p></li><li><p>What would you predict for 2000?</p></li><li><p>Average surface temperature compared with 1961-90 average (Source: Hadley Centre)</p></li><li><p>Causal modelling to take account of other variablesMultiple regression Very simple example at complex models E.g. the prediction of the demand for long term care for elderly people in 2031 at;Pos=1&amp;ColRank=1&amp;Rank=224The structure of the model is explained on page 2. It involves dividing the population into categories A key finding is that residential and nursing home places need to increase by 65%. Do you believe this will turn out to be right?</p></li><li><p>Causal modelling to take account of other variablesModels can get very complicated and use advanced mathsThis does not mean they are rightCommon sense should give a clue about what cannot be forecast!What factors are likely to be important for predicting the demand for long term care for elderly people in 2031?Are these incorporated in the model?</p></li><li><p>Are forecasts always right?Different forecasters produce different results e.g. see! They are almost never right. The question is how big is the error likely to be?Always consider the error in forecasts!</p></li><li><p>Things to check with forecastsCommon senseHistorical accuracy. How well did the methods do in the past? Measure by MAD, etc.Compare different forecastsAssumptions made (e.g. central assumptions)Probabilistic estimates may give a more realistic ideaLikely impact of chaos </p></li><li><p>ChaosIn theory, if we knewEverything about how the world works, andThe exact state of affairs nowWe should be able to forecast the future accurately for ever?Works for predicting positions of planets but not the weather, or human systemsOften small errors in (2) get magnified so predictions rapidly become useless this is called chaos. E.g. butterfly effect</p></li><li><p>RememberStatistics and forecasts are almost always less reliable than they may seem at first sight.Be careful!</p></li><li><p>When making or using forecasts of trends Remember the Things to check slidesRemember that rare, and so unpredictable, events may have a massive impact. These have been called black swans by Taleb (2008). Life unpredictable and only appears explicable in retrospect. Statistics based on past often misleading! Egs of black swans: the spread of the internet, the market crash of 1987, but not the present credit crunch which he says was predictable.</p></li><li><p>ReadingGordon (2008) especially Chapter 7 (see Unit Guide)Many books on standard statistical techniques. Also websites like if you want to check a particular techniqueTaleb, N. (2008). The black swan: the impact of the highly improbable. London: Penguin. (Part 2 is entitled We just cant predict. Econ and Stats as fraud.)Ayres, I. (2008). Super crunchers: how anything can be predicted. London: John Murray. (Lots of examples of impressive predictions using regression models)</p><p>*******************</p></li></ul>