 |
 |
Dear Mike, Thanks for your post. You have touched upon several points that could be usefully expanded. I'm particularly thinking of your separation of the roles of statistical methods and study design - and the need for both to be high quality and appropriately used. In a sense, ALL design activity consists of 'focused information gathering and analysis' to identify a design. In this, good "'focused information gathering and analysis' to identify a design" can be seen as the exact equivalent of good design process. This applies as equally to the most creative design activity as it does to designing (say pressure vessels) using tight highly-explicit international design standards. (As a mathematician I hate the redundancy of needing to write "'exact' equivalent", but from experience I know if I write 'equivalent' (which means '*exactly* the same') it will be interpreted instead as 'more or less similar') This focused information gathering and analysis that is the whole of design activity happens in many ways. For example, some happens inside individuals' heads, some happens semi-informally between people (e.g. brainstorming, crits, multidisciplinary design teams) , some happens using formal methods (e.g. statistical methods), some happens historically (previous findings are published), some happens in machines (e.g. in design software such as Photoshop and Solidworks). What you called 'good study design' is essential as it identifies the what and how of the information gathering and analysis. Statistical methods can play many different roles in fulfilling that 'what and how'. One of the problems in thinking about the roles of statistics in design, I suggest, is that the choice of roles of statistics tends to be strongly limited by the ways of thinking that come from two areas of traditional statistical application: manufacturing quality control and testing of hypotheses. These traditional perspectives can also be seen to apply to the style and limitations of the use of statistics in other fields (e.g. the social sciences, medicine and design research). There are other ways of using statistics (and mathematics in general) that apply more widely and more usefully - especially to design practice, and by implication to design research. A couple of examples: A simple combinatorics method used by Dr Sooyung Yang enable the exploration of garment shape from changing knit types in fixed needle count high fashion knitwear. I was the supervisor. More details of the method are available in http://www.lboro.ac.uk/microsites/lds/dprg-casestudies/ and in Yang, S and Love, T. (2009) Designing Shape-shifting of Knitwear by Stitch Shaping Combinatorics: A simple mathematical approach to developing knitwear silhouettes efficaciously. IASDR Conference 2009: Design / Rigor & Relevance, Seoul: International Association of Societies of Design Research and the Korean Society for Design [ www.love.com.au/docs/2009/SY&TL-IASDR.pdf ] Second, is from a study at the W-eB Research Centre in Perth of the design of micro-level supply chain management in retail in Japan. From memory, tiny but busy shops of the 7-11 variety were designed to carry minimal inventory and instead were supplied frequently by tiny vans delivering small amounts of goods. This 'just in time' design of supply of inventory aimed to have (say) a bottle of milk delivered at the back door of the shop just before a customer asked for a bottle of milk in the shop. The supply chain management system didn't know which customer would ask for the bottle of milk but from continually updating AI-based methods, it was predicted that there would be a call for a bottle of milk at that time. That was nearly 20 years ago. Things have moved on. Similar use of active statistics in design to the second example could be used in design in many other industries besides retail. One can see perhaps how the same approach might be used in graphic design and the design of visual outputs to align with or be ahead of ongoing changes in culture. There are intrinsically not real differences both in the variety of behaviour of individuals and groups between the two examples. More generally, I suggest there are more areas of design in which statistics can be used than the current traditional approaches. More importantly, there are more WAYS of using statistics and other mathematical methods and in more parts of design activity than those used at present. As you infer, this will require good skills both in statistical approaches and in study design, in consequence, a greater level of sophistication in theorising about design processes and practices. It will also require creative minds - even more so. Best wishes, Terry == Dr Terence Love MICA, PMACM, MAISA, FDRS, AMIMechE Director Design Out Crime & CPTED Centre Perth, Western Australia [log in to unmask] www.designoutcrime.org +61 (0)4 3497 5848 == ORCID 0000-0002-2436-7566 -----Original Message----- From: [log in to unmask] [mailto:[log in to unmask]] On Behalf Of Paul Mike Zender Sent: Tuesday, 16 January 2018 1:06 AM To: [log in to unmask] Cc: Paul Mike Zender <[log in to unmask]> Subject: Re: Inferential statistics applied to design Mauricio and Ricardo: Mauricio, good to hear from you too!! As often happens on the list, we can at times use the same words but mean different things. To clarify, I was thinking of inferential as opposed to descriptive statistics in my previous reply. And, I was thinking of statistics as a tool to analyze research data rather than as a stand-alone research method. Therefore, perhaps my assumptions were answering something a little different from what you were questioning. If so, I'm sorry. As your (Mauricio's) post suggests, using statistics (or anything else) to infer INDIVIDUAL PREFERENCES across populations is dangerous because personal preferences are inherently unstable. (Adapting Mauricio's analogy) I may want ham and cheese for lunch right now, but in 30 minutes I may have changed my mind and want soup instead. On the other hand, things like visual perception and human cognition are more generalizable across populations and making inferences based on those relatively stable operations should be more valid. In short, the STABILITY OF THE PHENOMENA is important to consider when making inferences. Some of your (Ricardo's) comments in your Jan 13 post suggest to me the importance of study design rather that the statistical means use to help understand the study. For example, good study design will be critical in determining whether "a choice occurred by chance or due to a deliberately proposed change by a designer." My takeaway here is a cliché: good statistics cannot make up for bad study design. I'm not saying that you described bad any study designs (other than perhaps your intentional reference to some of Tinker's work). GOOD STUDY DESIGN, one that accurately explores the hypothesis/question of interest, is an essential skill, a difficult and challenging skill I still work on with each study. I, and I infer (sorry, I couldn't resist) perhaps other design researchers as well, need better exposure to, and more skill at, the various methods you explained: cluster analysis; logistic regression; factorial analysis; and others. In our 2014 paper we used odds-ratio. I would not have known about "odds-ratio" if not for my colleague Amy. I'd welcome being at least a fringe part of a group of design researchers working to improve our field's appropriate use of statistical methods. An on-line course or consulting group are even possibilities. In my last post I mentioned Cincinnati Children's Hospital and the Research Foundation. They have a Division of people that provides statistical support to the other researchers. Advanced in research as their researchers are, they are not all equally skilled at statistical analysis and this Division helps insure appropriate methods are designed-into studies at the start. Perhaps design research would benefit from some group of experts or consultants on-line that could provide a similar service to design researchers worldwide. Ricardo, perhaps you are the one to organize this! Best... Mike Zender University of Cincinnati Visible Language ----------------------------------------------------------------- PhD-Design mailing list <[log in to unmask]> Discussion of PhD studies and related research in Design Subscribe or Unsubscribe at https://www.jiscmail.ac.uk/phd-design ----------------------------------------------------------------- ----------------------------------------------------------------- PhD-Design mailing list <[log in to unmask]> Discussion of PhD studies and related research in Design Subscribe or Unsubscribe at https://www.jiscmail.ac.uk/phd-design -----------------------------------------------------------------