Lawrence Moon
>Lawrence Moon
>Spinal Cord Injury >Stroke >Pathology and Repair
>Tools for Research >Experimental design and data handling >Laboratory Techniques
>How to get to the lab >Contact Lawrence


Experimental design and data handling

Here are the slides from my recent lab talk on Stats and Experimental Design... thanks to Dom Spina for letting me copy tons of his materials.

If you're a scientist or clinician making claims that (say) a drug treatment doesn't work, then you need to know how to handle negative data. You probably need to use confidence intervals: Altman, 2005.

How might you transform data to help satisfy the assumptions of ANOVA ? Check out Lew, 2007a.

If you do in vivo work, and you don't know about randomized block design and other types of factorial design, then you need to check out Michael Festing's website. And here's independent PROOF that you should follow his advice: Lew et al 2007b.

Randomized designs, of couse, need proper methods for randomization. Think you already know how? I did too, until I read Altman and Bland (1999).

In fact, Doug Altman has a whole set of excellent articles available.

Why are you using an outbred strain? In many cases, you could gain extra statistical power and learn about differences in genotype without using more animals. Check out examples.

Download two of Festing's peer-reviewed papers; Shaw et al., 2002 and Festing et al., 2002 and other related papers here. Essential reading.

If you think statistics is hell for scientists, then you need a guardian angel called Andy Field. Buy his book or use his website. You can test and improve your statistics using his (taxing) FLASH CARDS challenge or using his MULTIPLE CHOICE QUESTIONS. He also provides a set of resources for lecturers including powerpoints.

Statnotes is also excellent for SPSS users. It has a wealth of useful information about univariate and multivariate analysis.

For power analysis, check this out. I use G*Power 3 for calculation of samples sizes and power analysis.

For more on how to analyse experiments with a repeated measures design, see mixed models.

Quick Links

Nature Protocols manuscript and links

Publications as PDFs

Media coverage as PDFs