統(tǒng)計(jì)語(yǔ)言學(xué)教案(全英文版)綜述課件

1、Statistics in LinguisticsSyllabusSyllabusSyllabus SyllabusAssessment30% classroom participation70% final examCourse Books李紹山，2001，語(yǔ)言研究中的統(tǒng)計(jì)學(xué)，西安：西安交通大學(xué)出版社。Butler, C. 1985. Statistics in Linguistics. New York: Basil Blackwell.Woods, A., Fletcher, P. & Hughes, A. 1986. Statistics in Language Studies. Ca

2、mbridge: Cambridge University Press.ReferencesBrown, J. D. 1988. Understanding Research in Second Language Learning. Cambridge: Cambridge University Press.Brown, J. D. 2002. Doing Second Language Research. Oxford: Oxford University Press.Hatch, E. & Lazaraton, A. 1991. The Research Manual: Design an

3、d Statistics for Applied Linguistics. New York: Newbury House Publishers.Hinton, P. 2004. Statistics Explained. London: Routledge.Muijs, D. 2004. Doing Quantitative Research in Education with SPSS. London: Sage Publications.Statistical SoftwareSPSS (Statistical Package for the Social Sciences)Excel1

4、. IntroductionNeed for statistics in linguisticsReasonsIn language study, many kinds of work require the collection and analysis of quantitative data. Statistics turns the raw data into meaningful information for decision-making.In terms of research methods, articles of an argumentative, or speculat

5、ive nature are not convincing.“個(gè)人經(jīng)驗(yàn)與看法”在四個(gè)刊物中的總體比例從1978年的32%下降到1997年的12%（高一虹 等，1992）。Some knowledge of statistics will make us understand academic articles better.量化研究呈上升趨勢(shì)，現(xiàn)代外語(yǔ)的量化研究在80年代中才出現(xiàn)，到90年代經(jīng)常在20%-30%之間，從1997年第三期改版以來(lái)，量化研究的比例達(dá)到50%；外語(yǔ)教學(xué)與研究在80年代基本上徘徊在10%上下，而90年代基本上在30%-40%之間。（選自高一虹等，1999）Reasons

6、2. Describing VariablesPopulations, Samples, and Random SamplingPopulation(總體）Any collection of entities, of whatever kind, that is the object of investigation (Butler, C. Statistics in Linguistics, 1985)finite population: the number of entities is fixed and countable（有限總體）infinite population: the n

7、umber of entities is potentially infinite，at least in theory（無(wú)限總體）Sample （樣本）Entities selected from a population for investigation Populations, Samples, and Random SamplingSampling （抽樣）The process of selecting/drawing samples （樣本） from the population concernedReasons for sampling 1)For infinite popu

8、lations, it is impossible to make exhaustive investigations; for finite populations in which the number of entities is too large, exhaustive investigations are theoretically possible but impracticable. 2) Sampling cuts down the labor, time and cost involved in obtaining data. 3) Sampling minimizes t

9、he errors that can be easily made during the processing of large numbers of statistical data.By sampling, we hope that the results obtained from the sample will be generalisable to the population. 總體、樣本和隨機(jī)抽樣總體作為研究對(duì)象的任何個(gè)體的集合或目標(biāo)群體（李紹山，2008）樣本從總體中抽取的用以研究的一部分個(gè)體隨機(jī)抽樣抽簽隨機(jī)數(shù)表計(jì)算機(jī)生成的隨機(jī)數(shù)字 系統(tǒng)隨機(jī)抽樣（或準(zhǔn)隨機(jī)抽樣） 團(tuán)體抽樣分層隨

10、機(jī)抽樣： 比例分層隨機(jī)抽樣、非比例分層隨機(jī)抽樣多級(jí)抽樣總體抽樣框架隨機(jī)抽樣Populations, Samples, and Random Samplingrandom sampling(隨機(jī)抽樣): every unit (entity) in the population has an equal chance of being represented in the sample(總體中的每個(gè)個(gè)體有同等的機(jī)會(huì)在樣本中得到體現(xiàn)，即總體中的每個(gè)個(gè)體被抽中的概率是相同的。）drawing: simple random samplingrandom digits tablerandom digit

11、s generated by computer (Excel) Variation: systematic /quasi-random sampling Variation within systematic sampling: block samplingstratified random sampling proportional stratified random sampling disproportional stratified random samplingMulti-stage samplingPopulationSampling frameRandom samplingPop

12、ulations, Samples, and Random SamplingRandom samplingVariation of simple random sampling: systematic /quasi-random samplingThe first unit in the sample is selected by truly random methods, and then the other units are taken at equal intervals throughout the numbered population, the intervals being t

13、aken to give the desired number in the final sample. (interval=number of population/size of sample)Not truly random, since the second and subsequent units in the sample are not selected truly independently of the first unit. This does not matter seriously if there is no periodicity in the population

14、.Populations, Samples, and Random SamplingRandom samplingVariation of simple random sampling: systematic /quasi-random samplingVariation within systematic sampling: block sampling The position of the beginning of the sample is determined randomly, but the next N items are taken, where N is the size

15、of the sample.The position of the beginning of the sample is determined randomly, but the next N items are taken, where N is the size of the sample.The true randomness of such a sample is not guaranteed.It is important to realize that the selection of a sample by methods designed to achieve true ran

16、domness does not guarantee that the sample arrived at will indeed be representative of the population.Probability (Butler, C. Statistics in Linguistics, 1985 :7) Populations, Samples, and Random SamplingPopulations, Samples, and Random SamplingRandom samplingstratified random sampling: stratum-strat

17、a-stratify If the various subgroups (strata) in the population are known, then random sampling can be undertaken with each stratum, and the resulting subsamples can be combined to give an overall portional stratified random sampling: The proportion of each subsample in the overall sample i

18、s equal to the proportion of each stratum in the population. It enables a stratum that has a small overall proportion in the population to be represented in the sample and allows comparisons to be made between the subsamples. disproportional stratified random sampling: The proportion of each subsamp

19、le in the overall sample is not equal to the proportion of each stratum in the population. It provides the optimal condition for comparisons to be made. Populations, Samples, and Random SamplingRandom samplingmulti-stage sampling: The sampling is undertaken by successive stages, treating the sample

20、at one stage as the population for the next stage 目前我國(guó)高校英語(yǔ)專業(yè)包括專科在內(nèi)有1000多個(gè)辦學(xué)點(diǎn)最多的每年招收1000多名學(xué)生（王金生，朱黎輝，2008：45），其中本科英語(yǔ)專業(yè)教學(xué)點(diǎn)600多個(gè)，這其中有200多個(gè)設(shè)置在理工類院校（秦秀白，2006）。Populations, Samples, and Random Sampling Population parameter(參數(shù)） vs. sample statistic（統(tǒng)計(jì)量） parameter: one of the properties of a population, sym

21、bolized by Roman letters or English letters (李紹山, 2008) statisitc: one of the properties of a sample, symbolized by Greek lettersPopulations, Samples, and Random Samplingpopulation parametersample statistic/estimate(估計(jì)值）meanvariancestandard deviationcorrelation總體、樣本和隨機(jī)抽樣總體參數(shù) vs. 樣本統(tǒng)計(jì)量 Population Sam

22、ple Parameter Statistic平均數(shù) 方差 2 s2標(biāo)準(zhǔn)差 s相關(guān)系數(shù) rHomework 11. Define the following terms with an example. 1) population 2) sample 2. Why should we be content with a sample for our study in most cases?3. What is random sampling? Why should we select a sample randomly?4. 假如一所學(xué)校有15個(gè)系，每個(gè)系由20個(gè)自然班，每個(gè)班有25人左右，共

23、7300人。要從中選取一個(gè)100人的樣本，那么： 1）最好用什么方法進(jìn)行抽樣？為什么？請(qǐng)說(shuō)出具體步驟。 2）如果要保證樣本中男、女生各占一半，又如何進(jìn)行抽樣？ 3）試用隨機(jī)數(shù)表以簡(jiǎn)單隨機(jī)抽樣的方式從7300名學(xué)生中選取 該樣本。VariablesVariable ： an attribute (屬性） of a person, a piece of text, or an object, which varies from person to person, text to text, object to object, or from time to time.In statistics,

24、variables refer to measurable attributes, as these typically vary over time or between individuals. (From Wikipedia)VariablesIn a research project, we may wish to look at levels(水平） within a variable. For example, we might want to know how well ESL students are able to do some task. If the study is

25、designed to compare the performance of ESL students who are foreign students with the performance of those who are immigrant students, then the variable is the circumstance in which the ESL students learn English, and it will have two levels. If the study is concerned with geographic area, i.e. if t

26、he study is to compare the performance of students from different areas, then the variable is geographic area, and there might be such levels as South American, European, Middle Eastern, or Asian so that comparisons among these levels of ESL student can be made. The variable would consist of four le

27、vels. Or, for the purposes of the study, we might want to know if there is a difference in performance between advanced, intermediate, and beginner ESL students. The variable, then, is the proficiency level, which has three levels.The variable is ESL student. That variable may be divided into levels

28、 for the purposes of the study.(?)Variables Classification of variablesaccording to the function a variable has in a study independent variable & dependent variable according to the level of measurement nominal variable, ordinal variable, interval variable & ration variableAccording to whether data

29、obtained on a variable is continuous or not continuous variable & discrete (discontinuous) variableVariablesclassification of variablesaccording to the function a variable has in a study independent variable & dependent variableindependent variable(自變量）: variable that the investigator deliberately m

30、anipulates/ variable that the investigator can vary - IVdependent Variable（因變量）: variable whose response to the IV the investigator is measuring-DV Variablesclassification of variablesaccording to the level of measurement Nominal variable(名稱變量/稱名變量): purely qualitative, not quantitative. Entities ma

31、y be the same or different, but not “more” or “l(fā)ess”.Naming, not measurement, no arithmetic valuesex, social status, mother tongue ,marital status according to the level of measurement Ordinal variable(順序變量): rank the values of the entities on a scale of “more” or “l(fā)ess”. The value of one entity may

32、 be more or less, higher or lower than that of another entity.Ordering, cant tell the size of the differenceVery impoliteimpolitepolitevery politeThe lessons are boring 1 2 3 4 5 6 7 8 9 have arithmetic value, but the value is not preciseClass rank：1289Variableclassification of variablesaccording to

33、 the level of measurementInterval variable(等距變量）There are truly equal intervals between points on the scale of “more” or “l(fā)ess”.Test scores：95907065Temperatures:Ratio variable（比率變量）: there is an absolute zero in the variable, and ratios may be taken.Have an absolute zero point: height, time, distanc

34、e, but not temperature(絕對(duì)零度-273.15攝氏度)Seldom used in linguistic studiesVariablesClassification of variables according to the level of measurement It is very important for investigators to know which type of variable they are dealing with, because different statistical procedures are appropriate for

35、different types of variable. The level of measurement of a variable influences the choice of a measure of central tendency and variability, and influences the choice of procedures for hypothesis testing. Parametric tests are used for ratio and interval variables; non-parametric tests are used for no

36、minal and ordinal variables.英語(yǔ)期末考試分?jǐn)?shù)數(shù)據(jù)的變化：Variable IntervalOrdinalnominalStudents rankgroupingTest scoreHigh group, low groupranksVariablesClassification of variablesAccording to whether data obtained on a variable is continuous or not Continuous variable(連續(xù)變量）: may take any value within a given rangeFor a continuous variable, there are an infinite number of possible values that can fall between any two observed values. It is divisible into an infinite number of fractional parts, e.g. test scores, time spent on uttering a sentence, peopl