Yhteystiedot (FI)

K. Uusitalo, Home page

- The Empirical Cumulative Distribution Function, its Inaccuracy and Probability Plotting (text + program code, 2004). This article covers the theory of the ecdf and its transformation to make it suitable for probability plots. It appears, there are theoretically unjustified plotting position formulas in wide usage. However, while the widely used plotting position formulas give only one position per observation, the theory of order statistics reveals that the whole inaccuracy information related to the plotting positions can be presented by transformed beta distributions. The included program code can be used to draw the ecdf, normal quantile-quantile plots and Gumbel (return level) plots so that this inaccuracy information is presented. An animated example from the appendix is shown here: 100 x 30 Monte Carlo generated Gumbel(50,20) variates with median and confidence interval estimates of cumulative probability and corresponding return period.
- Distance Weighted Spatial Methods  (presentation, 2002). This presentation shows some tricks related to inverse distance weighted interpolation (IDW) as well as its theory. In my understanding, the traditionally used IDW formula is not theoretically grounded and the large number of far away data points affect too much on the local interpolation result if large search radius is used. For this reason I derived an IDW formula which takes into account, how the number of data points increases with distance. In addition the weight attenuation is considered to be proportional to data point density so that nearer data points shade the visibility of behind lying data points. Please notice that there are critical problems in most methods to compare different IDW formulas or to estimate IDW parameter values. The presentation shows also how to apply IDW to categorical data and how to take background variabels into account by using distance deformation. Unfortunately, there isn't explained that in most cases another method would be a better way to introduce background variables into IDW.


- Solantie R. & Uusitalo K. (2000): Patoturvallisuuden mitoitussadannat. Suomen suurimpien 1, 5 ja 14 vrk:n piste- ja aluesadantojen analysointi vuodet 1959-1998 kattavasta aineistosta. (Design precipitation for dam safety in Finland: An analysis of the largest 1-, 5- and 14-day point and areal precipitation totals during 1959-1998). Finnish Meteorological Institute. Reports 2000:3, 77 pp. (In Finnish)
- Tuomenvirta H., Uusitalo K., Vehviläinen B. & Carter T. (2000): Ilmastonmuutos, mitoitussadanta ja patoturvallisuus: Arvio sadannan ja sen ääriarvojen sekä lämpötilan muutoksista Suomessa vuoteen 2100. (Climate change, design precipitation and dam safety: Estimate of changes in precipitation, its extremes and temperature in Finland up to 2100). Finnish Meteorological Institute. Reports 2000:4, 65 pp. (In Finnish). This paper contains also a methodological innovation about extreme value analysis of sub-grid scale phenomena, see pp. 34-37. The central idea is that if rainfall events cover only a fraction of the grid cell local return levels and return periods are larger than grid cell return levels and return periods, i.e. the grid cell return level (the precipitation amount) divided by the fraction and (approximately) the grid cell return period divided by the fraction. It is a recommended excercise to check this on a Gumbel plot.


Metsänraja ja holoseenikauden ilmastokehitys Pohjois-Fennoskandiassa. (Forest line and climatic variations in northern Fennoscandia during Holocene). Helsingin yliopiston maantieteen laitoksen tutkimusretkiraportteja 41, 2005, pp. 18-31 (Seminar report for a university course in Finnish). Despite the title (which has been abbreviated by the serial editors), this paper covers in addition to the Holocene developments also the current state as I saw it during our expedition in northern Fennoscandia in late summer 2004.


Männistö S., Uusitalo K., Roos E., Fogelholm M. & Pietinen P. (1997): Alcohol beverage drinking, diet and body mass index in a cross-sectional survey. European Journal of Clinical Nutrition vol. 51, n:o 5, pp. 326-332. (My contribution to this paper was on the statistical analyses - covariance analysis, adjusted means).


A Shortwave Loop Antenna and its Enhancement by Positive Feedback. From this experiment I learned that positive (regenerative) feedback can cause a very huge signal amplification. This is something I wish all bureaucrats in their offices would ponder very deeply, because this has implications on how our society and organisations should work to be economically effective.


Contact: uusitalo at iki fi. Please do not publish my contact information without my consent. Web pages: The relocation address http://www.iki.fi/uusitalo/ points to this page and it will most probably remain valid even in case the actual location of this page happened to change. My coordinates: I live in Helsinki, Finland.