My WebLink
|
Help
|
About
|
Sign Out
Home
Browse
Search
FLOOD01900
CWCB
>
Floodplain Documents
>
Backfile
>
1001-2000
>
FLOOD01900
Metadata
Thumbnails
Annotations
Entry Properties
Last modified
11/23/2009 10:40:54 AM
Creation date
10/4/2006 10:26:47 PM
Metadata
Fields
Template:
Floodplain Documents
County
Statewide
Community
Nationwide
Title
Techniques of Weather-Resources Investigations of the USGS Book 4, Chapter A1
Date
1/1/1968
Floodplain - Doc Type
Educational/Technical/Reference Information
There are no annotations on this page.
Document management portal powered by Laserfiche WebLink 9 © 1998-2015
Laserfiche.
All rights reserved.
/
44
PDF
Print
Pages to print
Enter page numbers and/or page ranges separated by commas. For example, 1,3,5-12.
After downloading, print the document using a PDF reader (e.g. Adobe Reader).
Show annotations
View images
View plain text
<br />6 <br /> <br />TECHNIQUES OF WATER-RESOURCES INVESTIGATIONS <br /> <br />. <br /> <br /> <br />o <br /> <br />x <br /> <br />Figure 7.-Hypothetical sampling distribution of means. <br /> <br />latter condition the probability is small of <br />obt~ing an X of such size from a population <br />havmg a mean of zero. Therefore, we would <br />reject the hypothesis and would state that the <br />result was ~ignificant at a certain probability <br />level, meanmg that the results obtained differ <br />significantly from the hypothesis. <br />A common problem is the test of significance <br />?f a r~gression coefficient. The null hypothesis <br />IS aga~n t~at the true value of the regression <br />coeffiCient IS zero, and the test may be made in <br />the same way as before. However, the pro- <br />cedure commonly used is somewhat different. <br />The confidence limits about the theoretical <br />value are computed. If the regression coefficient <br />is b, its standard error S" and its population <br />value 13, then the limits are found to be <br /> <br />b-tS,<13<b+tS,. <br /> <br />where t is the appropriate value for the chosen <br />significance level and sample size. If the limits <br />!nclude zero, the hypothesis is accepted, that <br />IS, the regression coefficient is not significantly <br />different from zero. If the limies are both on <br />one side of zero, the hypothesis is rejected and <br />t~e regressi?n coefficient is considered sig- <br />nific~ntly different from zero, that is, it is <br />considered meaningful. <br /> <br />Many other tests of significance are available <br />but all parametric tests are based on the theor; <br />of sampling and follow the general procedure <br />described above. A less powerful group of non- <br />pa;~met?c .test~ may be used when the prob- <br />ability distributiOn of the statistic is not known. <br />(See Siegel, 1956.) <br /> <br />Correlation and Regression <br /> <br />The distinctions between correlation and <br />regression must be recognized in order to apply <br /> <br />and interpret either of the methods. These <br />distinctions are very marked although they may <br />seem of little'importance because of the simi- <br />larity of the computation proqedures. Dixon <br />and Massey (1957, p. 189) made the following <br />distinction between the two: <br />"A regression problem considers the fre- <br />quency distribution of one variable when <br />another is held fixed at each of several levels. A <br />correlation problem considers the joint varia- <br />tion of two measurements, neither of which is <br />restricted by the experiment." <br />Correlation is a process by which the degree <br />of association between samples of two variables <br />is defined. The correlation coefficient is a <br />mathematical definition of that association. <br />It is, of course, possible to compute a correlation <br />coefficient from any two sets of data. The <br />mathematical definition of association implies <br />no cause-and-effect relation nor even that the <br />relation between the two variables results <br />from a common cause. <br />Correlation theory requires that the data be <br />drawn randomly from a bivariate normal dis- <br />tribution. However, McDonald (1957) reported <br />that experimental sampling studies show the <br />nonnormality effects, usually regarded as dis- <br />turbing by statisticians, to be of incousequential <br />magnitude geophysically. A further require- <br />ment of correlation is that both variables X <br />and Y be without error due to measurement. <br />Nothing can be measured without error, so the <br />above requirement is one of degree. The <br />question of the error allowable is subject to <br />arbitrary decisions, particularly since the true <br />error of the data is never known. <br />The end product of the process of correlation <br />is the correlation coefficient; it is not an equa- <br />tion. The equations which describe Y as a <br />function of X, and X as a function of Y, are <br />regression equations, not correlation equations. <br />Another way of stating the distinction between <br />correlation and regression is that correlation <br />measures the degree of association between two <br />variables, whereas regression prmrides equations <br />for estimating individual values of one variable <br />from given values of the other. <br />Reliability of correlation results depends on <br />the number of items used to compute the <br />correlation coefficient and the mllgnitude of the <br />computed correlation coefficient. Confidence <br /> <br />. <br /> <br />. <br />
The URL can be used to link to this page
Your browser does not support the video tag.