330 likes | 345 Views
Understanding Standards Event. Higher Statistics Award. Hypothesis Testing David Young Department of Mathematics & Statistics University of Strathclyde NHS, Greater Glasgow and Clyde. Overview. Welcome to the University of Strathclyde i ntroduction to SQA level 6 Statistics Award
E N D
Understanding Standards Event Higher Statistics Award
Hypothesis Testing David Young Department of Mathematics & Statistics University of Strathclyde NHS, Greater Glasgow and Clyde
Overview Welcome to the University of Strathclyde • introduction to SQA level 6 Statistics Award (SQA Home > National Qualifications > Subjects > Statistics Award (SCQF level 6) • how statistics can answer important questions • practical sessions using Minitab and/or R
What is statistics …? • statisticsis the science of collecting, analysing, presenting and interpreting data • it enables the objective evaluation of research questions of interest • it provides the means to weigh up how much evidence the collected data provide for and against the research hypothesis of interest • statistics are available on almost every aspect of daily life
2014 – the year of the selfie • millions are shared each and every day across all the major social media platforms • according to data from Samsung, selfies make up almost one-third of all photos taken by people aged 18-24 • Facebook, WhatsApp, Twitter, Instagram and Snapchat bearing the brunt of that load • Link – http://www.adweek.com/digital/selfie-generation/ • Twitter’s most retweeted post of all time … • Twitter’s most retweeted post of all time • …
Addressing important questions • Can you get diabetes from eating too much sugar? • What are the side effects of too much vitamin D? • Is smartphone addiction a real phenomenon? • Does eating turkey at Christmas make you tired? • Does vitamin C make you look younger? • Is paracetamol better than codeine for treatment of pain? • ‘Two heads are better than one’. • Statistics can be used to answer such questions
Not always a clear answer … • turkey makes you tired because it’s loaded with tryptophan • tryptophan is a pre-cursor to serotonin which is a mood modulator, inducing relaxation and supporting sleep • tryptophan alone does not make you tired • combining tryptophan with carbohydrates does • it is not the turkey that’s causing Christmas sleepiness, it’s the turkey plus the cranberries, rolls, and pie • turkey actually has less tryptophan than chicken
Drug development • pharmaceutical companies have to gather data on drug performance • a drug cannot be used commercially until the company have produced evidence of efficacy and safety • statistical tests (known as hypothesis test) can be used to gather this evidence
Omeprazole • Omeprazole (Prilosec, Zegerid) belongs to a group of drugs called proton pump inhibitors • it decreases the amount of acid produced in the stomach • Omeprazole is used to treat symptoms of gastroesophageal reflux disease(GERD) and other conditions caused by excess stomach acid • common side effects may include: stomach pain, gas, nausea, vomiting, diarrhoea or headache • https://www.drugs.com/omeprazole.html
Statistics and Probability • statistical analysis considers the probability of an event being due to chance • can never be 100% certain for example that one treatment is better than another • can say mathematically how sure we are that a result is true
Probability • hypothesis: all dogs can bark • how would you prove this? • data has to be gathered to give evidence for or against the hypothesis • a hypothesis test has to be performed
Interpretation • If the difference was not due to chance why did it happen?
Interpretation • If the difference was not due to chance why did it happen? • When is this claim valid?
Interpretation • If the difference was not due to chance why did it happen? • When is this claim valid? • What are the consequences of getting it wrong?
Additional Points • errors in hypothesis testing – p<0.01! • null and alternative hypotheses • cranberry juice – randomisation http://www.ncbi.nlm.nih.gov/pubmed/22961092
Additional Points • double blind studies • http://www.theguardian.com/society/2005/jan/17/health.medicineandhealth • placebo trials • comparison of baseline characteristics • intention-to-treat and per-protocol – weight loss example • tests for correlation, regression and normality testing
Two heads are better than one …? • study design? • what were the rules for doing the quiz? • randomisation? • double blind? • how many participated – why is this important? • who marked? • are you satisfied that the groups were comparable?