This session given by Dr Ian Handel from the Roslin Institute in the UK was great fun. Handel asked us what we wanted to cover in the session and wrote them on a whiteboard, crossing them off as we went. We played some fun games with die and chickpeas which funnily enough, didn’t work out the way he wanted it to go first off so we started again – resulting in ‘proper’ results. Tee hee! It was a fast-paced session which might have made some game players count wrongly (me amongst them). So, did I learn anything new? I do have some rudimentary knowledge of statistics but thought I should attend this session because I find I have to have it repeated for me since I don’t use the knowledge often. I tend to forget. Statistics can be dry and for people without a firm mathematical footing, confusing. Handel was very enthusiastic which made the session fun and the really good thing was that he started at the beginning – with mean median and mode. Once you get these down, the rest becomes a little easier to follow. The standard deviation is the range or how spread out your sample is. The null hypothesis (this is something I didn’t get in my rushed introduction to statistics) is the underpinning of determining Type 1 and Type 2 errors and the basis for thinking about p values. A null hypothesis is that there is nothing going on or that there this no effect. It is the opposite of your hypothesis. I found a good video recently that describes type I and type II errors. The p value is set before the study is done and is a % indicating significance. And here is a handy mnemonic – if the p is low, the null must go! Big samples better pin down the differences between groups, but the groups have to be similar – you have to compare like with like. Big samples equal smaller confidence intervals (estimates of population parameters) while smaller samples have larger confidence intervals. We didn’t go into great detail about parameters so I might have to go over these again later on.
Handel discussed how visual depictions of statistics can be misleading, using a bar chart created by some US government department using very large values (something like 100|1000|10000|15000) on the Y axis. This made the items measured look equal but when you reconfigured the Y axis measurements to a more reasonable configuration (100|500|1000|1500) the equality became seriously inequality. He recommends 3 books: The Tiger That Isn’t: Seeing Through a World of Numbers | The Visual Display of Quantitative Information | Dicing with Death: Chance, Risk and Health. He also recommends 2 websites – one is about funny correlations eg: the number of drownings by falling into pools correlates with the number of films Nicholas Cage starred in and BBC Radio’s statistics program, More or Less.
Are there any shortcuts in creating search strategies to collect information for systematic reviews? Wichor Bramer advocates using MS Word macros to translate strategies from one database to another. We had a go at creating one during this workshop but – and this is the only downside to Wichor and Gerdien’s workshop – there wasn’t enough time to experiment. Once you get it done, it ends up as a search paragraph that only needs minor editing. For example, if you start in EMBASE first, you have to add TI and AB for Medline in EBSCOHost. Another example is if you have adjacency numbers, you have to edit them according to what interface you are using (NEAR/3 in EMBASE is really 2 words whereas in EBSCOHost N3 is 3 words) etc. So much easier than copying and pasting it line by line. How do you do it? For ease of process, always start with a single database (they recommend EMBASE.com). Then in a word doc, write ( ) and within the parenthesis ‘ ‘ /exp, then your desired thesaurus term and the staying within the parenthesis, ( ):ab,ti . It will take me a lot of practice! Full details in the handout: [Improving efficiency & confidence in systematic literature searching]. There is a slight error in the handout though – DE is only used in PsycINFO for non-major subjects. All other subjects in EBSCOHost use MH. Wichor’s team have made the macros they use available online. Download and installation instructions are given in the handout.
This isn’t advocating speed (though you do speed up naturally the longer your acquaintances with databases) – care and thought still needs to be put into developing the search strategy. I learnt some useful tips: use truncation in PubMed MeSH database (I don’t know why I didn’t think of trying this!), not out individual thesaurus terms per element to see if any relevant articles appear (and check what thesaurus terms relevant articles use) – vice versa for title/abstract terms (also scan relevant article abstracts for terms not used in the title/abstract), and use floating subheadings with thesaurus terms to increase specificity. Wichor also mentioned that including adjectives in the search query could introduce bias eg harsh parenting. I hadn’t thought of it in that way before, but I have thought that using them, especially in the Outcomes section of PICO, may restrict results unnecessarily. How many people use those sorts of terms in an abstract?
I’m glad I was able to attend this workshop – thank you Wichor and the team at Erasmus MC.
30.06.15 Update: Wichor has updated the handout and the corrected version is here.
This is great
How many times have you ended up with an aching back after continuously bending to read the bottom of posters at conferences? This is not a problem I’ve had but I realised after this session that I avoided looking at posters for too long to avoid this very problem. And that is what this session had as an icebreaker – what problems do you have with posters? This icebreaker focused on design problems though – and there are lots of these too. How do you fit all the information you want to have on your poster? Where do you put graphics and text? What about colours, fonts, graphs etc? Why create one anyway?
Look at the title of this one
There are lots of good things about posters. You can revisit them and you don’t have to talk to the creator if you don’t want to. If you are a creator, it is a good way to share research to a wide audience. They can transcend language barriers, are good at presenting ideas in different ways and you can also have a printout for the viewer to take away. It can act as a teaser to a future paper and if you have one already or a website, screencast or podcast, you can add a QR code to it so the viewer could, if they wanted to, go straight to that for more information.
A good one ..
A good poster is one with inclusive design. A good poster is easy for dyslexics and people with colour or vision disabilities to make sense of, easy for people with bad backs or knees to read (don’t put text on the bottom!), and has a reading flow that makes it easy for people who read right-left and left-right to navigate. There are some basic design tips to remember: use sans serif fonts – (and don’t mix them!), the header should be as large as possible for contrast with rest of the text, punchy headings draw in the audience, balance space and information, visuals should complement the text, don’t use acronyms unless the audience knows what it refers to (I guess MeSH is a good example of an acronym lots of medical librarians recognise whereas RMH, the acronym of my workplace, is not), bar charts are more immediately understandable than pie charts, and make sure your poster is understandable from a distance. Finally – remember to follow the conference poster guidelines!
Better but for the title size
Our attempt …
We all got the chance to feel like kids again when we worked in groups creating a poster (cutting and pasting pictures from magazines onto butcher paper) and had fun with the mini poster presentations. We all got a great handout [Create a Great Poster] with tips and were pointed to some useful resources to use when choosing palettes, contrast, and designing materials to suit people who are visual learners as well as people who are text readers. I’ve only created one poster (years ago now) and after this session, I am keen to try my hand at another.