The New Statistics
by Emeritus Professor Geoff Cumming
To listen to an audio podcast, mouse over the title and click Play. Open iTunes to download and subscribe to iTunes U collections.
These ten videos introduce descriptive statistics, then sampling, confidence intervals, and meta-analysis. They explain why estimation and meta-analysis ('The New Statistics') are much better than traditional null hypothesis significance testing (NHST). Enjoy the 'Dance of the p values', which highlights just how unreliable p values are. Confidence intervals are much more informative. There is currently a shift occurring from NHST to The New Statistics. This is a great development, which will lead to improved research progress.
|1||VideoEpisode 1: Frequency Distributions||A frequency histogram and a dot plot prompt good appreciation of a data set. The ESCI software makes it easy to explore such data pictures.||10/6/2013||Free||View In iTunes|
|2||VideoEpisode 2: Descriptive Statistics||ESCI presents the mean, median, and standard deviation, using pictures to give good understanding.||10/6/2013||Free||View In iTunes|
|3||VideoEpisode 3: The Normal Distribution||See the beautiful bell-shaped picture of the normal distribution, and explore using ESCI.||10/6/2013||Free||View In iTunes|
|4||VideoEpisode 4: Sampling||ESCI simulations introduce random sampling from a normal population. Enjoy the dance of the means.||10/6/2013||Free||View In iTunes|
|5||VideoEpisode 5: Standard Error||ESCI simulations illustrate that the standard error is the standard deviation of the sampling distribution of the sample mean.||10/6/2013||Free||View In iTunes|
|6||VideoEpisode 6: Estimation||Enjoy the dance of the confidence intervals, which is the basis of statistical estimation.||10/6/2013||Free||View In iTunes|
|7||VideoEpisode 7: Confidence Intervals||Confidence intervals are highly informative, and give the best basis for interpreting research results.||10/6/2013||Free||View In iTunes|
|8||VideoEpisode 8: NHST to Estimation||Traditional null hypothesis significance testing (NHST) has severe problems. Statistical cognition provides evidence that it is much better to use estimation and not NHST.||10/6/2013||Free||View In iTunes|
|9||VideoEpisode 9: Dance of the p Values||ESCI simulations demonstrate the enormous variability of the p value. Don't trust p! The dance of the p values is enormously wide, simply because of sampling variability. Estimation is way better than NHST.||10/6/2013||Free||View In iTunes|
|10||VideoEpisode 10: Meta-Analysis||Meta-analysis is the integration of evidence from a number of studies on the same or similar questions. See the forest plot, which is a beautiful picture that summarises a meta-analysis.||10/6/2013||Free||View In iTunes|