From ARarig@health.state.ak.us Tue Jun 16 11:19:16 1998 Received: from mxu3.u.washington.edu (mxu3.u.washington.edu [140.142.33.7]) by lists.u.washington.edu (8.8.4+UW97.07/8.8.4+UW97.05) with ESMTP id LAA36860 for ; Tue, 16 Jun 1998 11:19:15 -0700 Received: from fy-jnu1x.health.state.ak.us (fy-jnu1x.health.state.ak.us [146.63.178.72]) by mxu3.u.washington.edu (8.8.4+UW97.07/8.8.4+UW97.09) with ESMTP id LAA06237 for ; Tue, 16 Jun 1998 11:19:12 -0700 Message-Id: <199806161819.LAA06237@mxu3.u.washington.edu> Received: by fy-jnu1x.health.state.ak.us with Internet Mail Service (5.5.1960.3) id ; Tue, 16 Jun 1998 10:19:08 -0800 From: "Rarig, Alice" To: waphgis@u.washington.edu Subject: RE: Census data for assessment Date: Tue, 16 Jun 1998 10:12:59 -0800 MIME-Version: 1.0 X-Mailer: Internet Mail Service (5.5.1960.3) Content-Type: text/plain Showing the counts on a map is a great idea (although it can be challenging -- we're planning to use objects that vary in diameter with the count to help show it visually -- e.g. firearm injuries -- and may have to switch scales for the cities) -- at least as part of the picture -- I think showing counts AND rates, with good footnotes explaining the need for care in interpreting rates, can then be helpful. Counts themselves can provide too little info if people are unfamiliar with the pop distributions. For example, two fire deaths a year average over ten years in a rural county with 5000 people would be a whole lot more alarming than the same average in a county of 200,000. > ---------- > From: Hoskins, Richard E.[SMTP:reh0303@hub.doh.wa.gov] > Sent: Tuesday, June 16, 1998 9:40 AM > To: waphgis@u.washington.edu > Subject: RE: Census data for assessment > > Dr Rarig's comments are really on the mark, and especially > " state the source for the denominator so readers can > make their own judgments about how that affects the output." > > I would add that for rate calculations for inference it is important > to > calculate a confidence interval; most formulas for which assume that > the > denominator has > no variance which is clearly not true. Any CI you calculate would be > wider if the variance of the denominator were taken into account. It > is > tempting to calculate a disease rate in a county for several years and > > then look at the confidence interval overlap to get some idea about > statistically significant change from one year to the next or even > just > compare with a regional or state value - but if the population model > supplies no estimate of the variability of the population estimate > these > trends will not be possible to detect or quantify. And if the numbers > are > small anyway ... I wonder if it might be best to just display count > data on maps and leave it at that .. forget rates. > > Richard E. Hoskins, PhD MPH > WA State Public Health Geographer > GIS and Spatial Epidemiology > Office of Epidemiology > WA State Department of Health > 1102 Quince St, Olympia, WA 98504-7812 > > eMAIL: REH0303@hub.doh.wa.gov > tel: 360-236-4270 > fax:360-236-4245 > > .