Abstract Some Issues In Recognizing Causal Relationships

Some Issues In Recognizing Causal Relationships

Lawrence Mazlack

Sarah Coppock

Computer Science

University of Cincinnati

Cincinnati, Ohio 45220

{Mazlack,Coppock}@https://www.wendangku.net/doc/da15770319.html,

Abstract

Determining causality has been a tantalizing goal throughout human history. Proper sacrifices to the gods were thought to bring rewards; failure to make the proper observations were thought to lead to dis-aster. Today, data mining holds the promise of extracting unsuspected information from very large data-bases. The most common build association rules from large data sets.Association rules indicate the strength of association of two or more data attributes. In many ways, the interest in association rules is that they offer the promise (or illusion) of causal, or at least, predictive relationships. However, association rules only calculate a joint occurrence frequency; they do not express a causal relationship. If causal rela-tionships could be discouvered, it would be very useful. Our goal is to explore causality in the data mining context.

Key Words: causality, data mining, association rules, imbalanced causality

1 Introduction

Recognizing when causality occurs implies recognizing a causal relationship; e.g., A causes B to happen. Whether this can be done at all has been a speculation for thousands of years. At the same time, in our daily lives, we make the common sense observation that causality exists. Carrying this common sense observation further, our concern is how to computationally recognize a causal rela-tionship. Our concern is the discouvery of causal relationships in large data sets.

Causal relationships exist in the common sense world. If someone fails to stop at a red light and there is an automobile accident, we say that the per-son’s failure to stop is the cause of the accident. Another way we think of causal relationships is in a counterfactual sense. For example, if the driver dies in the accident we might say that had the accident not occurred; they would still be alive.

Data mining holds the promise of extracting unsuspected information from very large data-bases. The most common build association rules from large data sets. Association rules indicate the strength of association of two or more data attrib-utes. In many ways, the interest in association rules is that they offer the promise (or illusion) of causal, or at least, predictive relationships. Whether any association rules express a causal relationship needs to be examined.

Causality occupies a position of centrality in human reasoning. In particular, it plays an essential role in human decision-making by providing a ba-sis for choosing that action that is likely to lead to a desired result.

The ability to form association rules leads to the intuitive desire to infer causal relationships among the items (Mazlack, 2001). A knowledge of causal relationship would be a useful data mining product. However, association rules only describe a joint frequency of the co-occurrence of attribute values. As typically formed, association rules do not represent causal relationships. However, asso-ciation rules might be useful as a first step in causal discouvery. A discouvered association might flag a potentially interesting relationship; then a causal discovery method might test the relationship.

1.1 Controlled Data

A common approach to recognizing causal re-lationships is by manipulating a variable through experimentation while observing another variable.

Although there are developed techniques to discouver causal relationships among controlled1

1Controlled data means that the actions producing particular values can be reproduced. Typically, this is experimental data.

data, how to do so in purely observational2 data is not solved. Observational data is the most likely to be available for analysis; especially in potential data mining applications.

Real world causal events are often affected by a large number of potential factors. For example, with the growth of a plant, many factors such as air temperature, chemicals in the soil, types of crea-tures present, etc., can all affect plant growth. What is unknown is what causal factors will be present in the data; and, how many of the underlying causal relationships can be discouvered among pure ob-servational data. Taking into account that the data was not collected for the sole purpose of deter-mining causality, it is likely that some factors inter-acting with the variables will not be present in the data.

In the existing algorithms, assumptions are made in order to infer the relationships. The as-sumptions often concern the nature of the data distribution. In addition, statistical testing is often used and therefore, the problem of adequate sam-ple (e.g., size of the data set) and selection bias exists.

Some problems with discouvering causality include:

?Defining adequately a causal relation,?Representing possible causal relations, and ?Infer causes and effects from the representation.

1.2 Discouvery In Observational Data

The present algorithms for discouvery in ob-servational data often use correlation and probabil-ity independence to find possible causal relation-ships. For example, if two variables are statistically independent, it can be asserted that they are not causally related.

One of the limiting factors in being able to in-fer causal relationships in mining is the need to use observed data rather than controlled data. Another difficulty is that the analysis might be post facto;

i.e., after the data has been collected without an opportunity participate in the selection of the re-corded variables. Of course, control can be exerted over the attributes selected for consideration from amongst the already recorded data.

Existing algorithms make assumptions and lack some characteristics that may be desirable such as strength of the causal relationship. The issues of adequate sample and selection bias exist.

2Observational data indicates that the data has been produced and no repeated actions can be taken reproduce further data.

Two important approaches to discouvering causality in observational data, are: constraint-based and Bayesian. The current algorithms do not address any strength or likelihood of the causal relationships inferred. For example, it may be dis-couvered: a causes b and c causes d. Assuming no interrelationships between the two sets, can we say that a’s relationship to b is stronger or weaker than c’s relationship to d? If we are using the knowl-edge for prediction, the strength of the relationship would affect the judgment of prediction.

2Representation

An open and important issue is the representa-tion chosen. Representation is important as it con-strains the discouvery method that can be used. Some representations for use in discouvering cau-sality have been proposed. The representations al-low for the characterization and inference of causal relationships. Some representations include:?Digraphs, such as, directed ascyclic graphs (DAGs), (Sprites, 2000) (Pearl, 2000),?Probability trees (Shafer, 1998), and

? First-order logic (Hobbs, 2000).

2.1 Digraphs

In a digraph, the vertices correspond to the variables and each directed edge from v

1

to v

2

cor-responds to a causal influence from v

1

to v

2

. This is shown in Figure 1, where both gender and parent’s education have a causal influence on education. Pearl (2000) and Sprites (2000) use a form of di-graphs called DAGs for representing causal rela-tionships. A DAG consists of a set of vertices, V, and a set of directed edges between the vertices, e(v

1

,v

2

) where v

1

,v

2

are in V. As a representation for causal relationships, the vertices correspond to the variables (or attributes) that are in a data set. Each directed edge from v

1

to v

2

corresponds to a causal influence from v

1

to v

2

. Figure 1(a) shows an ex-ample of a possible digraph representing causal relationships.

gender

parent's

education

education

(b)Example of variables that follow the

structure in (a)

Figure 1(b) represents both gender and parent’s education having a causal influence on a person’s education. According to this representation, gender and parent’s education also have a causal influence

on a person’s salary even though the relationship is through education.

Sometimes, cycles exist. For example, a per-son’s family medical history influences both whether they are depressive and whether they will have some diseases. Drinking alcohol combined with the genetic predisposition to certain diseases influences whether the person has a particular dis-ease, which then influences pain, which thereby may influence the person’s drinking habits. Figure 2 shows the cyclic digraph for this example.

Figure 2. Example of cyclic causal relationships 2.2 Probability Trees

Probability trees can be used to show causal in-fluences Shafer (1998). A tree is a digraph starting from one vertice, the root. In this case, the vertices represent situations. Each edge represents a par-ticular variable with a corresponding probability.

Time ordering of the variables is represented via the levels in the tree. The higher a variable is in the tree, the earlier it is in time. This can become ambiguous for networked representations; i.e., when a node can have more than two parents and thus two competing paths (and their imbedded time sequences).

By evaluating the expectation and probability changes among the situations (or node) in the tree, one can decide whether the two variables are caus-ally related or not.

2.3 First Order Logic

Hobbs (2000) illustrates the use of first-order logic to represent causal relationships. The predi-cate “causal complex” is defined to indicate a group of events that obtain an effect. This seems a more suitable representation for inference than a DAG or probability tree. One difficulty with this approach is that the representation does not allow for any gray areas. For example, if the event that may have occurred when the wind was blowing east, how could the wind’s blowing east-northeast be accounted for? The causality inferred may be incorrect due to the rigidity of the representation. Also, there is no description of the strength of the relationship.

Nor can first order logic deal with dependen-cies that are only sometimes true. For example, sometimes when the wind blows hard, a tree falls. This kind of sometimes event description can pos-sibly be statistically described. Alternatively, a qualitative fuzzy measure might be applied.

Another problem is recognizing differing ef-fect strengths. For example, if it is found that a specific causal complex is discouvered causing a particular event, then is it true that some events in the causal complex are more strongly tied to the effect? Also, it is not clear how a relationship such as the following would be represented: a causes b some of the time; b causes a some of the time; other times there is no causal relationship.

3 Nature Of Causal Relationships

Algorithms for discouvering association rules in data with the same observational characteristics have been developed. Since we are able to discou-ver association rules, the question becomes: can we use this to infer to any degree causal relationships? This would still be valuable to discouver, even if we find that there is no causal relationship.

Some questions about causal relationships that would be desirable to answer are:

?To what degree does a cause b? Is the value for b sensitive to a small change in the value for a??Does the relationship always hold in time and in every situation? If it does not hold, can the par-ticular situation when it does hold be discouvered?

?Is it possible that there might be mutual depend-encies; i.e., a → b as well as b → a? Is it also pos-sible that they do so with different strengths?

When speaking of mutual dependencies, we might describe them as shown in Figure 3.

a

b

Figure 3. Mutual dependency notation.

where S

i,j

represents the strength of the causal rela-tionship from i to j . For example, a could be short men and b could be tall women. If S

a,b

meant the romantic attraction that was caused in short men by the sight of tall women, we might discouver that S

a,b

> S

b,a

It would seem that if there are causal relation-ships in market basket data, there would often be imbalanced dependencies. For example, if a cus-tomer first buys strawberries, there is a reasonably good chance that she will then buy whipped cream. Conversely, if she first buys whipped cream, the subsequent purchase of strawberries is probably less likely.

The present algorithms for discouvery in ob-servational data use correlation between variables and probabilistic independence to find possible causal relationships. If two variables are discou-vered to be independent, it is more than likely the two are not causally related. On the other hand, when two variables are found to be correlated, how can it be decided whether one causes the other?

It is potentially interesting to discouver the ab-sence of a causal relationship. For example, assume it is found that shelf position in a food store (e.g., top shelf, bottom shelf, etc.) does not cause a change in the sale of Cheerios, but shelf position (e.g. middle shelf) does cause a change in the sale of Special K. Then, the decision maker can use this information to adjust their shelves accordingly. A more important use for discouvering lack of a causal relationship is in discouvering causes of dis-ease. If some potential cause can be eliminated, then attention can become more focused on other potentials.

4Discouvery Based On Graphical Representations

Two approaches to discouvering causality in observational data are constraint-based and Bayesian. Some algorithms do not address the strength or likelihood of the causal relationships inferred. For example, if it may be discouvered a causes b and c causes d, can we say that a’s rela-tionship to b is stronger or weaker than c’s rela-tionship to d? If we are using the knowledge for decision-making, the strength of the relationship would affect the decision-making.

4.1 Constraint-Based Discouvery

Using a graphical representation (Pearl, 2000) (Sprites, 2000) (and the underlying algorithms) assumes that the variables satisfy the Markov con-dition. The Markov condition in this instance means that for every vertice, v, given the values of the parents of v, v is independent of all other verti-ces v' that are not descendants of v. In the case of causality, it may be reasonable to assume this. But, this assumption eliminates the potential discouvery of cyclic causal relationships. Because cycles may exist in the real world, they could exist in the data.

It is possible for underlying cycles to exist. For example, a person’s family medical history influ-ences both whether they are depressive and whether they will have some diseases. Drinking alcohol combined with the genetic predisposition to certain diseases influences whether the person has a par-ticular disease, which then influences pain, which may therefore influence the person’s drinking habits. Figure 2 shows the digraph for this exam-ple. Although cycles may exist in the real world, we cannot assume that they have been captured in the data. If the data set includes all of the variables in-volved in the cycle, then it will exist. On the other hand, there may be an attribute involved in the cy-cle missing from the data.

4.2 Reductive Discouvery

A reductive discouvery algorithm contains all possible edges and then selectively eliminates some of them. The PC (Spirtes, 2000) and the IC (Pearl, 2000) algorithms derive a possible underly-ing DAG by constraining the edges in the graph. For each pair of vertices, the smallest set of vertices on which the two variables become independent is computed. This information is used to either add an edge or remove an edge.

In the PC algorithm, an edge is removed if the two vertices are d-separated given a non-empty set of vertices. Two vertices x & y are said to be d-separated given a set of vertices Z if Z blocks every path from x to y. Z is said to block a path p be-tween x and y if

? p contains a chain, i→j→k or i ? j ? k, or a fork, i←j→k where j is in Z; or

? p contains an inverted fork, i→j←k, such that j is not in Z and no descendent of j is in Z.

Edges are then oriented according to the sets of nodes that separate each pair of nodes. D-separation is a graphical way of expressing condi-tional independence among variables.

4.3 Additive Discouvery

Additive discovery is the complement of re-ductive discovery. It starts with nothing and adds DAGs as their need is discouvered

The IC algorithm begins with no edges. If no set of vertices is found that causes an unconditional independence for two vertices, then an edge is added between the two vertices. In other words, the two variables are always dependent with the given variables. After adding all undirected edges as pos-sible, the edges are then oriented according to the sets found in the previous step. In other words, di-rection is added to the edges according to set of given variables for which the two vertices become independent. The neighboring edges also play a role in which direction the arrow is added between the two vertices.

Both the IC (Pearl, 2000) and the PC (Spirtes,2000) algorithm have been modified to take into account latent (or unmeasured) variables. Again,this assumption of missing variables is a necessary assumption. The resulting graphs are more compli-cated. Rather than a resulting partially oriented graph, a partially oriented and partially marked graph results.

4.4 LCD Based Algorithms

The approach of Silverstein (1998) for mining causal structures builds on the LCD algorithm (Cooper, 1997) while retaining its polynomial time complexity. These algorithms do not attempt to discouver the complete causal structure as the IC and PC algorithms do. The LCD algorithm finds causal structures in the form of chains and forks.Without added complexity, an additional causal structure is found; the authors term this structure CCU causality. CCU causality is represented by the structure a → b ← c , and is also referred to as v-structures (Pearl, 2000).

The authors use support 3. This restricts the testing of causality to items that would be of more interest. A confidence threshold value is used as the confidence level of the statistic used in determining whether two variables are dependent. The chi-squared (χ2) statistic is used for determining de-pendence of two variables. ?2 is the statistic com-puted as: (O -E )/E where O is the observed value for the variables and E is the expected value for the variables. If ?2 is greater than χ2α, where á is the confidence level of the test, then the two variables are dependent. Using this statistic in combination with support and confidence, the error of deter-mining dependence is reduced.

Silverstein (1998) doesn’t assume that in the case a → b , a is a direct cause. This allows for pos-sible hidden variables. Other assumptions are made that would be desirable to eliminate. They assume only Boolean data with no missing data. For exam-ple, this method could be used on Boolean data extracted from market basket data.

By eliminating the use of graphical representa-tion, the approach is more computationally feasi-ble. But, it does not overcome the difficulty in identifying structure with a certain amount of as-surance. Lastly, these methods only assume certain statistical correlation’s in a particular orientation among the variables to decide if one variable causes another. Usually, the time ordering of the variables is assumed known beforehand. When the time order is not known, Pearl (2000) introduces the notion of statistical time, any variable ordering that coincides with the causal structure.

3 Support means that particular values must be above a

certain frequency in the data.

5Bayesian Causal Discouvery

A Bayesian approach as described by Hecker-man (1995) entails finding the most probable causal structure and the corresponding parameters for the structure. The first complexity issue en-countered is the number of possible models. For just 3 variables, there are 25 possible models.Clearly, it would be infeasible to enumerate all of the possible models for even a small data set. To help solve this problem, Heckerman (1995) men-tions different approaches have been proposed such as selecting only one “best” model accord-ing to the user’s knowledge and then computing the parameters. Selecting only a particular portion of the possible models and searching over the se-lected as though they are exhaustive is another option.

As with many data mining methods, this places an over emphasis on the use of a human expert.Even with an expert, bad choices might be made.Unimportant factors might be included; important factors ignored. A dependence on expert opinion also begs the question. If we are certain that we know the causal factors, there is little reason to run a computationally expensive discouvery algorithm.It is because we are unsure of the causal factors that we need to investigate a solution by using discou-very algorithms.

The Bayesian approach can allow for missing data. However, allowing missing data also increases the complexity issues even larger than previously.One other thing to note, this approach also assumes the nature of variables' distribution in addition to the difficulty of assessing priors to the models.Geiger (1995) shows that there is a known distri-bution that is suited to be used as priors given in-dependence of the parameters. This distribution is the Dirichlet distribution. Heckerman (1995) shows how to use this distribution to derive the priors for the parameters and to update the priors.

6 Epilogue

Causality occupies a central position in human common sense reasoning. In particular, it plays an essential role in human decision-making by pro-viding a basis for choosing that action that is likely to lead to a desired result.

Recognizing when causality occurs implies recognizing a causal relationship. Whether this can be done at all has been a speculation for thousands of years. At the same time, in our daily lives, we make the common sense observation that causality exists. Carrying this common sense observation further, our concern is how to computationally recognize a causal relationship. Our concern is the discouvery of causal relationships in large data sets.

Today, data mining holds the promise of ex-tracting unsuspected information from very large databases. Methods have been developed to build association rules from large data sets. Association rules indicate the strength of association of two or more data attributes. In many ways, the interest in association rules is that they offer the promise (or illusion) of causal, or at least, predictive relation-ships. However, association rules only calculate a joint probability; they do not express a causal rela-tionship. If causal relationships could be discou-vered, it would be very useful.

Discouvering causal relationships is of great interest. When making decisions, it is often useful to know the relationships between events or items so that this information can be taken into account. If it is found that a causes b, and b is not desired, then this becomes useful information. The interest in defining and discouvering causality is displayed by the amount of literary discussion in fields such as statistics, philosophy, social sciences and others.

Although there are techniques developed to discouver causal relationships among controlled data; how to do so in purely observational data is not solved. Graph based methods are excessively complex when faced by a typical large data set. Our final paper will explore these issues in greater depth.

We are particularly interested in determining when causality can be said to be stronger or weaker. Either in the case where the causal strength may be different in two independent relationships; or, where in the case where two items each have a causal relationship on the other.

To answer these questions, we also must con-sider how to: recognize when there is a causal rela-tionship, measure the causal strength, and represent the relationship. All of these are open issues on which we are working.

References

R. Agrawal, T. Imielinski, A. Swami [1993]“Mining Association Rules Between Sets Of Items In Large Databases,” Proceedings Of ACM SIGMOD Conference On Management Of Data (SIGMOD-93), 207-216

G. Cooper [1997] “A Simple Constraint-Based Algorithm for Efficiently Mining Observational For Causal Relationships” in Data Mining and Knowledge Discouvery, v 1, n 2, 203-224

D. Geiger, D. Heckerman [1995] “A Characteriza-tion Of The Dirichlet Distribution With Application To Learning Bayesian Networks,” in Proceedings of the 11th Conference on Uncertainty in AI, Mont-real, Quebec, 196-207, August C. Glymour, G. Cooper, eds. [1999] Computation, Causation, and Discouvery, AAAI Press, Menlo Park, California

D. Heckerman [1995] A Tutorial On Learning With Bayesian Networks, Microsoft Research Paper, MSR-TR-95-06.

L. Mazlack [2001] “Considering Causality In Data Mining,” WSES/IEEE Multi-Conference, Crete, 493-498

J. Pearl, J. [2000] Causality: Models, Reasoning, And Inference, Cambridge University Press NY, NY.

G. Shafer [1998] “Mathematical Foundations For Probability And Causality” Proceedings Of Sym-posia In Applied Mathematics, v 55, 207-270

C. Silverstein, S. Brin, et al. [1998] “Scaleable Techniques For Mining Causal Structures," Pro-ceedings 1998 International Conference Very Large Data Bases, NY, 594-605

P. Spirtes, C. Glymour, R. Scheines [2000] Causa-tion, Prediction, And Search, MIT. Cambridge Massachusetts.

初中英语介词用法归纳总结

初中英语介词用法归纳总结 常用介词基本用法辨析 表示方位的介词：in, to, on 1. in 表示在某地范围之内。 Shanghai is/lies in the east of China. 上海在中国的东部。 2. to 表示在某地范围之外。 Japan is/lies to the east of China. 日本位于中国的东面。 3. on 表示与某地相邻或接壤。 Mongolia is/lies on the north of China. 蒙古国位于中国北边。 表示计量的介词：at, for, by 1. at 表示“以……速度”“以……价格”。 It flies at about 900 kilometers an hour. 它以每小时900公里的速度飞行。 I sold my car at a high price. 我以高价出售了我的汽车。 2. for 表示“用……交换，以……为代价”。 He sold his car for 500 dollars. 他以五百元把车卖了。

注意：at表示单价(price) ，for表示总钱数。 3. by 表示“以……计”，后跟度量单位。 They paid him by the month. 他们按月给他计酬。 Here eggs are sold by weight. 在这里鸡蛋是按重量卖的。 表示材料的介词：of, from, in 1. of 成品仍可看出原料。 This box is made of paper. 这个盒子是纸做的。 2. from 成品已看不出原料。 Wine is made from grapes. 葡萄酒是葡萄酿成的。 3. in 表示用某种材料或语言。 Please fill in the form in pencil first. 请先用铅笔填写这个表格。They talk in English. 他们用英语交谈。 表示工具或手段的介词：by, with, on 1. by 用某种方式，多用于交通。 I went there by bus. 我坐公共汽车去那儿。 2. with表示“用某种工具”。

some和any的用法与练习题

some和 any 的用法及练习题( 一) 一、用法： some意思为：一些。可用来修饰可数名词和不可数名词，常常用于肯定句 . any 意思为：任何一些。它可以修饰可数名词和不可数名词，当修饰可数名词 时要用复数形式。常用于否定句和疑问句。 注意： 1、在表示请求和邀请时，some也可以用在疑问句中。 2、表示“任何”或“任何一个”时，也可以用在肯定句中。 3、和后没有名词时，用作代词，也可用作副词。 二、练习题： 1.There are ()newspapers on the table. 2.Is there ( )bread on the plate. 3.Are there () boats on the river? 4.---Do you have () brothers ?---Yes ,I have two brothers. 5.---Is there () tea in the cup? --- Yes,there is () tea in it ,but there isn’t milk. 6.I want to ask you() questions. 7.My little boy wants ()water to drink. 8.There are () tables in the room ,but there aren’t ( )chairs. 9.Would you like () milk? 10.Will you give me () paper? 复合不定代词的用法及练习 一.定义: 由 some，any，no，every 加上 -body,-one,-thing,-where构成的不定代词，叫做复合不定代词 . 二. 分类： 1.指人：含 -body 或 -one 的复合不定代词指人 . 2.含-thing 的复合不定代词指物。 3.含-where 的复合不定代词指地点。 三：复合不定代词： somebody =someone某人 something 某物，某事，某东西 somewhere在某处，到某处 anybody= anyone 任何人，无论谁 anything任何事物，无论何事，任何东西 anywhere 在任何地方 nobody=no one 无一人 nothing 无一物，没有任何东西 everybody =everyone每人，大家，人人 everything每一个事物，一切 everywhere 到处 , 处处 , 每一处

初中英语名词练习题与详解

名词 判断对错 1、［误］ Please give me a paper. ［正］ Please give me a piece of paper. ［析］不要认为可以数的名词就是可数名词，这种原因是对英语中可数与不可数名词的概念 与中文中的能数与不能数相混淆了，所以造成了这样的错误，因paper 在英语中是属于物质名词一类，是不可数名词。而不可数名词要表达数量时，要用与之相关的量词来表达，如： two pieces of paper. 2、［误］ Please give me two letter papers. ［正］ Please give me two pieces of letter paper. ［析］ paper 作为纸讲是不可数名词，而作为报纸、考卷、文章讲时则是可数名词，如：Each student should write a paper on what he has learnt. 3、［误］ My glasses is broken. ［正］ My glasses are broken. 4、［误］ I want to buy two shoes. ［正］ I want to buy two pairs of shoes. ［析］英语中glasses—眼镜， shoes—鞋， trousers—裤子等由两部分组成的名词一般要用复 数形式。如果要表示一副眼镜应用 a pair of glasses 而这时的谓语动词应与量词相一致。如：5、This pair of glasses is very good. ［误］ May I borrow two radioes? ［正］ May I borrow two radios? ［析］以o 结尾的名词大都是用加es 来表示其复数形式，但如果 o 前面是一个元音字母或外来语时则只加s 就可以了。这样的词有zoo— zoos,piano—pianos. 6、［误］ This is a Mary's dictionary. ［正］ This is Mary's dictionary. ［析］如名词前有指示代词this, that, these those, 及其他修饰词our,some, every, which,或所有格时，则不要再加冠词。 7、［误］ There are much people in the garden. ［正］ There are many people in the garden. ［析］可数名词前应用 many, few, a few, a lot of 来修饰，而 people 是可数名词，而且是复数名词，如： The people are planting trees here. 8、［误］ I want a few water. ［正］ I want a little water. ［析］不可数名词前可以用 a little, little, a lot of, some来修饰，但不可用many,few 来修饰。 9、［误］ Thank you very much. Y our family is very kind to me. ［正］ Thank you very much. Y our family are very kind to me. 10、［误］ Tom's and Mary's family are waiting for us. ［正］ Tom's and Mary's families are waiting for us. 11、［误］ I'm sorry . I have to go. Tom's families are waiting for me. ［正］ I'm sorry. I have to go. Tom's family are waiting for me. ［析］集合名词如果指某个集合的整体，则应视为单数，如指某个集合体中的个体则应视为 复数。如 :My family is a big family. When I came in, Tom's family were watching TV. 即汤姆一家人正在看电视。这样的集合名词有：family class, team 等。

初中英语介词用法总结

初中英语介词用法总结 介词(preposition)：也叫前置词。在英语里，它的搭配能力最强。但不能单独做句子成分需要和名词或代词（或相当于名词的其他词类、短语及从句）构成介词短语，才能在句中充当成分。 介词是一种虚词，不能独立充当句子成分，需与动词、形容词和名词搭配，才能在句子中充当成分。介词是用于名词或代词之前，表示词与词之间关系的词类，介词常与动词、形容词和名词搭配表示不同意义。介词短语中介词后接名词、代词或可以替代名词的词（如：动名词v-ing）.介词后的代词永远为宾格形式。介词的种类： （1）简单介词：about, across, after, against, among, around, at, before, behind, below, beside, but, by, down, during, for, from, in, of, on, over, near, round, since, to, under, up, with等等。 （2）合成介词：inside, into, outside, throughout, upon, without, within （3）短语介词：according to, along with, apart from, because of, in front of, in spite of, instead of, owing to, up to, with reguard to （4）分词介词：considering, reguarding, including, concerning 介词短语：构成 介词+名词We go to school from Monday to Saturday. 介词+代词Could you look for it instead of me? 介词+动名词He insisted on staying home. 介词+连接代/副词I was thinking of how we could get there. 介词+不定式/从句He gives us some advice on how to finish it. 介词的用法： 一、介词to的常见用法 1.动词+to a)动词+ to adjust to适应, attend to处理；照料, agree to赞同,

some和any的用法

some和any的用法： （1）两者修饰可数单数名词，表某一个；任何一个；修饰可数复数名词和不可数名词，表一些；有些。〔2）一般的用法：some用于肯定句；any用于疑问句，否定句或条件句。 I am looking for some matches. Do you have any matches? I do not have any matches. （3）特殊的用法： (A) 在期望对方肯定的回答时，问句也用some。 Will you lend me some money? (=Please lend me some money.) (B) any表任何或任何一个时，也可用于肯定句。 Come any day you like. （4）some和any后没有名词时，当做代名词，此外两者也可做副词。 Some of them are my students.〔代名词） Is your mother any better?（副词） 3. many和much的用法： （1）many修饰复数可数名词，表许多； much修饰不可数名词，表量或程度。 He has many friends, but few true ones. There hasn't been much good weather recently. （2）many a: many a和many同义，但语气比较强，并且要与单数名词及单数形动词连用。 Many a prisoner has been set free. (=Many prisoners have been set free.) （3）as many和so many均等于the same number of。前有as, like时, 只用so many。 These are not all the books I have. These are as many more upstairs.

some与any的用法区别教案资料

s o m e与a n y的用法 区别

some与any的用法区别 一、一般说来，some用于肯定句，any用于否定句和疑问句。例如： She wants some chalk. She doesn’t want any chalk. Here are some beautiful flowers for you. Here aren’t any beautiful flowers. 二、any可与not以外其他有否定含义的词连用，表达否定概念。例如： He never had any regular schooling. In no case should any such idea be allowed to spread unchecked. The young accountant seldom (rarely, hardly, scarcely) makes any error in his books. I can answer your questions without any hesitation. 三、any可以用于表达疑问概念的条件句中。例如： If you are looking for any stamps, you can find them in my drawer. If there are any good apples in the shop, bring me two pounds of them. If you have any trouble, please let me know. 四、在下列场合，some也可用于疑问句。 1、说话人认为对方的答复将是肯定的。例如： Are you expecting some visitors this afternoon?（说话人认为下午有人要求，所以用some）

50套初中英语数词

50套初中英语数词 一、初中英语数词 1．We throw rubbish every year. A. ton of B. tons C. tons of D. a ton of 【答案】 C 【解析】【分析】句意：我们每年扔大量的垃圾。ton，吨，前面没有具体数字，因此用tons of，大量的，故选C。 【点评】考查固定搭配，注意平时识记。 2．Two students to the opening ceremony last Friday. A. hundreds; were invited B. hundred; were invited C. hundreds of; invited 【答案】 B 【解析】【分析】句意：上周五有200名学生被邀请参加开幕式。根据题干中的two与选项中的hundred可知此题考查确切数量的表达方式，hundred要用单数形式；students与invite存在动宾关系，此处要用被动语态，由last Friday，可知要用一般过去时，故选B。【点评】考查数量的表达方式以及被动语态。注意确切数量与不确切数量在表达上的不同。 3．—Do you know the boy is sitting next to Peter？ —Yes. He is Peter's friend. They are celebrating his birthday. A. who; ninth B. that; nine C. which; ninth 【答案】 A 【解析】【分析】句意：——你知道那个坐在彼得旁边的男孩吗?——是的。他是彼得的朋友。他们正在庆祝他的九岁生日。分析句子结构可知，第一空所在句子是定语从句，先行词是人，连接词在从句中作主语，所以应该用who/that引导，which连接定语从句时先行词应该是物，故排除C；nine九，基数词；ninth第九，序数词；第二空根据空后的birthday为名词单数可知，此处需要序数词，表示某人几岁生日应该用序数词表示第几个生日，故选A。 【点评】考查定语从句的连接词的辨析和序数词。注意区别定语从句的连接词的使用原则，理解单词词义。 4．There are ________________ months in a year. My birthday is in the ________________ month. A. twelve; twelve B. twelfth; twelfth C. twelve; twelfth D. twelfth; twelve 【答案】 C 【解析】【分析】句意：一年有12个月，我的生日在第12个月。名词复数months前是基数词，twelve是基数词，the定冠词后是序数词，twelfth是序数词，故选C。 【点评】考查数词，注意名词复数前是基数词，定冠词后是序数词的用法。

some和any的用法

some和any的用法 1.some adj.一些；某些；某个pron. 某些；若干；某些人 a.adj. some可以修饰可数名词或不可数名词，意为“某些”。 Some people are playing football. (some+可数名词) I ate some bread. (some+不可数名词) b.adj. some后面可以修饰可数名词的单数，意为“某(个)”。 Some day you will know. (some+可数名词的单数) 有一天你会知道的。 Some student cheated in the exam.(some+可数名词的单数) 有个学生考试作弊。 对比：Some students cheated in the exam.有些学生考试作弊。 c.pron. some此时作代词，后面不需要再加名词就可以表示“有些(人)”的意思。 All students are in the classroom, and some are doing their homework. d.pron. some作代词，意为“若干(…)”。 There are 10 apples on the table. You can take some. 桌上10个苹果，你可以拿走一些。 2.any adj.任何的；所有的pron.任何一个；任何 a.adj. any可以修饰可数名词和不可数名词，意为“任何的，所有的任何一（…）”。 （用于否定意义的陈述句、疑问句、条件状语从句if中） Do you have any ideas?(any+可数名词复数)(疑问句) 你有什么想法吗？ I don’t have any bread.(any+不可数名词)(否定意义的陈述句) Please tell me if you have any problem.(if引导的条件状语从句) b.any后面可以加可数名词的单数，意为“任何一（…）”。 Any error would lead to failure.(any+可数名词单数) 任何（一个）错误都会导致失败。 c.pron. any此时作代词，与some里面c点的用法相似，只是表示这个意义的时候，any多用于否定句和疑问 句中。 比较：There are 10 apples on the table. You can take some. 桌子上有10个苹果，你可以拿走一些。 There are 10 apples on the table, but you can’t take any. 桌子上有10个苹果，但是你不能拿。 There are some apples on the table.桌上有些苹果。 There aren’t any apples on the table.桌上没有苹果。 由此，把陈述句变为否定句/一般疑问句的时候，要把some改成any。 思考：some只用于肯定句，any只用于否定句和疑问句中吗吗？ 不一定，要看句子本身想表达的意思。 1.some可以用于肯定句和疑问句中。在表示请求、邀请、提建议等带有委婉语气的疑问句中，用some表示说话 人希望得到肯定的回答。例如： Would you like some coffee?你想喝咖啡吗? 这里用some而不用any，是因为说话人期待得到对方肯定的回答。 （因此Would you like…?你想要…吗?这个句型中多用some而不用any） 比较： Do you have any books?这里用any而不用some，说明这只是因为这只是纯粹的疑问。

初中英语最全英语介词用法

表示方位的介词：in, to, on 1. in 表示在某地范围之内。 如：Shanghai is/lies in the east of China.上海在中国的东部。 2. to 表示在某地范围之外。 如：Japan is/lies to the east of China. 日本位于中国的东面。 3. on 表示与某地相邻或接壤。 如：Mongolia is/lies on the north of China.蒙古国位于中国北边。 表示计量的介词：at, for, by 1. at表示“以……速度”“以……价格” 如：It flies at about 900 kilometers a hour.它以每小时900公里的速度飞行。 I sold my car at a high price. 我以高价出售了我的汽车。 2. for表示“用……交换，以……为代价”如：He sold his car for 500 dollars. 他以五百元把车卖了。 注意：at表示单价(price) ，for表示总钱数。

3. by表示“以……计”，后跟度量单位 如：They paid him by the month. 他们按月给他计酬。 Here eggs are sold by weight. 在这里鸡蛋是按重量卖的。 表示材料的介词：of, from, in 1. of成品仍可看出原料 如：This box is made of paper. 这个盒子是纸做的。 2. from成品已看不出原料 如：Wine is made from grapes. 葡萄酒是葡萄酿成的。 3. in 表示用某种材料或语言 如：Please fill in the form in pencil first. 请先用铅笔填写这个表格。 They talk in English. 他们用英语交谈。 注意：in指用材料，不用冠词；而with指用工具，要用冠词。如：draw in pencil/draw with a pencil 表示工具或手段的介词：by, with, on 1、by用某种方式，多用于交通 如by bus乘公共汽车，by e-mail. 通过电子邮件。

人教版初中英语短语大全最全

人教版初中英语短语大全 1)be back/in/out 回来/在家/外出 2)be at home/work 在家/上班 3)be good at 善于，擅长于 4)be careful of 当心，注意，仔细 5)be covered with 被……复盖 6)be ready for 为……作好准备 7)be surprised (at) 对……感到惊讶 8)be interested in 对……感到举 9)be born 出生 10)be on 在进行，在上演，（灯）亮着 11)be able to do sth. 能够做…… 12)be afraid of (to do sth. that…)害怕……(不敢做……，恐怕……) 13)be angry with sb. 生(某人)的气 14)be pleased (with) 对……感到高兴(满意) 15)be famous for 以……而著名 16)be strict in (with) (对工作、对人)严格要求 17)be from 来自……，什么地方人 18)be hungry/thirsty/tired 饿了/渴了/累了 19)be worried 担忧 20)be (well) worth doing (非常)值得做…… 21)be covered with 被……所覆盖…… 22)be in (great) need of (很)需要 23)be in trouble 处于困境中 24)be glad to do sth. 很高兴做…… 25)be late for ……迟到 26)be made of (from) 由……制成 27)be satisfied with 对……感到满意 28)be free 空闲的，有空 29)be (ill) in bed 卧病在床 30)be busy doing (with) 忙于做……(忙于……) (二)由come、do、get、give、go、have、help、keep、make、looke、put、set、send、take、turn、play等动词构成的词组 1)come back 回来 2)come down 下来 3)come in 进入，进来 4)come on 快，走吧，跟我来 5)come out出来 6)come out of 从……出来 7)come up 上来 8)come from 来自…… 9)do one's lessons/homework 做功课/回家作业 10)do more speaking/reading 多做口头练习/朗读 11)do one's best 尽力 12)do some shopping (cooking reading, clean ing)买东西(做饭菜，读点书，大扫除) 13)do a good deed (good deeds)做一件好事(做好事) 14)do morning exercises 做早操 15)do eye exercises 做眼保健操 16)do well in 在……某方面干得好 17)get up 起身 18)get everything ready 把一切都准备好 19)get ready for (=be ready for) 为……作好准备 20)get on (well) with 与……相处(融洽) 21)get back 返回 22)get rid of 除掉，去除 23)get in 进入，收集 24)get on/off 上/下车 25)get to 到达 26)get there 到达那里 27)give sb. a call 给……打电话 28)give a talk 作报告 29)give a lecture (a piano concert)作讲座(举行钢琴音乐会) 30)give back 归还，送回 31)give……some advice on 给……一些忠告 32)give lessons to 给……上课 33)give in 屈服 34)give up 放弃 35)give sb. a chance 给……一次机会 36)give a message to……给……一个口信 37)go ahead 先走，向前走，去吧，干吧 38)go to the cinema 看电影 39)go go bed 睡觉(make the bed 整理床铺) 40)go to school (college) 上学(上大学) 41)go to (the) hospital 去医院看病

初中英语介词用法归纳整理

初中英语介词用法归纳整理 表示时间的介词 at：用于表示时刻，时间的某一点。 on:用于星期，某天，某一天的上午，下午，晚上指具体的某一天时，一律用on in:用于表示周，月，季节，年，泛指上午，下午，晚上 before:在...之前 after:在...之后 by:在....前时间截止到... untiltill：直到.....为止 for：达...之久表示过了多少时间 during：在....期间 through：一直..从开始到结束 from：从...起时间 since:自从...以来表示从以前某时一直到现在仍在继续 in：过...后未来时间 within：不超过...的范围 表示场所，方向的介词 at :在某地点表示比较狭窄的场所 in:在某地表示比较宽敞的场所 on:在...上面，有接触面 above：在...上方 over：在...正上方，是under的反义词 under：在..下面，在...之内 below :在...下方不一定是正下方

near：近的，不远的 by:在...的旁边，比near的距离要近 between：在两者之间 among：在三者或者更多的之中 around：环绕，在...的周围，在....的四周 in front of:在...的前面 behind：在...后边 in：在..之内，用于表示静止的位置 into：进入 out of :和into一样，也表示有一定的运动方向 along：沿着 across：横过平面物体 through：贯通，通过 to :达到..地点目的地或方向 for：表示目的，为了..... from：从...地点起 其他介词 with：和..在一起; 具有，带有; 用某种工具或方法 in：表示用什么材料例如：墨水，铅笔等或用什么语言。表示衣着.声调特点时，不用with而用in。 by:通过...方法，手段 of:属于...的，表示...的数量或种类 from：来自某地，某人，以...起始 without：没有，是with的反义词 like :像...一样

some和any地用法

(1)some和any 的用法： some一般用于肯定句中,意思是“几个”、“一些”、“某个”作定语时可修饰可数名词或不可数名词。如：I have some work to do today. (今天我有些事情要做)/ They will go there some day.(他们有朝一日会去那儿) some 用于疑问句时,表示建议、请求或希望得到肯定回答。如：Would you like some coffee with sugar?(你要加糖的咖啡吗？) any 一般用于疑问句或否定句中,意思是“任何一些”、“任何一个”,作定语时可修饰可数或不可数名词。如：They didn’t have any friends here. (他们在这里没有朋友)/ Have you got any questions to ask?(你有问题要问吗？) any 用于肯定句时,意思是“任何的”。Come here with any friend.(随便带什么朋友来吧。) (2)no和none的用法： no是形容词,只能作定语表示,意思是“没有”,修饰可数名词(单数或复数)或不可数名词。如：There is no time left. Please hurry up.(没有时间了,请快点) / They had no reading books to lend.(他们没有阅读用书可以出借) none只能独立使用,在句子中可作主语、宾语和表语,意思是“没有一个人(或事物)”,表示复数或单数。如：None of them is/are in the

classroom.(他们当中没有一个在教室里) / I have many books, but none is interesting.(我有很多的书,但没有一本是有趣的) (3)all和both的用法： all指三者或三者以上的人或物,用来代替或修饰可数名词；也可用来代替或修饰不可数名词。 both指两个人或物,用来代替或修饰可数名词。all和both在句子中作主语、宾语、表语、定语等。如：I know all of the four British students in their school.(他们学校里四个英国学生我全认识) / --Would you like this one or that one? –Both.(你要这个还是那个？两个都要。) all和both既可以修饰名词(all/both+(the)+名词),也可以独立使用,采用“all/both + of the +名词(复数)”的形式,其中的of 可以省略。如：All (of) (the) boys are naughty.(是男孩都调皮) (4)every和each用法： every是形容词,只能作定语修饰单数名词,意思是“每一个”,表示整体概念； each是形容词、代词,可用作主语、宾语、定语等,意思是“每个”或者“各个”,表示单个概念；each可以放在名词前,可以后跟of 短语,与动词同时出现时要放在“be动词、助动词、情态动词”之后或者行为动词之前 every和each都用作单数理解,但是下文中既可以用单数的代词(如he/him/his)也可以用复数的代词(如they/them/their)替代。如：Every one of the students in his class studies very hard.(他班上每个学生学

some,any,one ones those that的区别和用法

some和any的区别和用法 要表示"一些"的意思，可用some, any。 some 是肯定词，常用于肯定句；any是非肯定词，常用于否定句或疑问句。例如： There are some letters for me. There aren''t any letters for me. Are there any letters for me? I seldom get any sleep these days. any也常用于条件分句以及带有否定含义的句子中： If you have any trouble, please let me know. 如果你有任何麻烦，请让我知道。 I forgot to ask for any change. 我忘了要一些零钱。 当说话人期待肯定回答时，some也可用于疑问句, 比如当说话人期待来信时，他可以问道：Are there some letters for me? 当购物时向售货员提问或者主人向客人表示款待时，也可在疑问句中用some: Could I have some of these apples? some和any 既可以修饰可数名词又可以修饰不可数名词，some常用在肯定句中，而any则常用在否定和疑问句中。因此 some和any 的用法主要是考虑用在肯定句、疑问句还是否定句中，与名词的可数与否无关。 some意为“一些”，可作形容词和代词。它常修饰可数名词复数。如：some books一些书，some boys一些男孩，也可修饰不可数名词，如：some water一些水，some tea一些茶叶，some常用在肯定句中。any意为“任何一些”，它也可修饰可数名词复数或不可数名词，常用于疑问句和否定句。如： --I have some tea here. 我这儿有些茶叶。 --I can’t see any tea. 我没看见茶叶。 --Do you have any friends at school? 你在学校有些朋友吗? --I have some English books, they are my best friends. 我有英语书，它们是我最好的朋友。 但在表示建议，反问，请求的疑问句中，或期望得到肯定回答时，多用some而不用any。如：Would you like some coffee? 你要不要来点咖啡? What about some fruit juice? 来点水果汁如何? 当any表示“任何”的意义，起强调作用时，它可以用在肯定句中； Any student can answer this question.任何学生都可以回答这个问题。 选题角度： 辨析some和any的不同用法：some 常用在肯定句中，而any 则常用在否定和疑问句中。在表示建议，反问，请求的疑问句中，或期望得到肯定回答时，多用some而不用any。一般用于疑问句或否定句中，用于never, hardly, without等词之后，用于if / whether 之后。而some则用于肯定句中，用于建议或请求的疑问句中，用于预料会作肯定回答的疑问句中，用于表示反问的否定的疑句中。 如： 1. I’d been expecting ________ letters the whole morning, but there weren’t ________ for me. (全国卷)卷 A. some; any B. many; a few C. some; one D. a few; none

英语量词用法大全

英语量词用法大全 常用语法 场景1：当和外国友人走在公园，突然看到一群鸽子。这时候你是不是就会说:There are many pigeons 或者a lot of呢？想再高大上一点吗？Sam, look! There is a flock of pigeons over there. 场景2:当说三张纸的时候，是不是很习惯说：“three papers”？错了哦。paper 是不可数名词。正确说法是：three pieces of paper。 ◆◆量词用法整理+讲解◆◆ 量词可以用于描述可数名词比如a herd of elephants(一群大象）, 也可以描述不可数名词；比如three pieces of paper（三张纸） 1、描述一群...量词+群+of+名词 一群人a/an crowd／group／army／team／of people； 一群牛、象、马、天鹅a herd of cattle／elephants／horses／swans

一群鸟、鹅、母鸡、羊、燕子a flock of birds／geese／hens／goats／swallows 一群猎狗、狼a pack of hounds／wolves 2、描述一丝／点／层 一丝怀疑a shadow of doubt 一线未来之光a glimpse of future 一缕月光a streak of moonlight 一层霜／雪／糖霜a layer of frost／snow／cream 3. piece块；片；段；项；件；篇；首；幅；张 a piece of bread／paper／wood／furniture／land／advice／news／meat ／cloth／music... 4、英译“一阵” 一阵哭泣／喝彩／炮击／ 雷声

some与any的用法区别

some与any的用法区别 一、一般说来，some用于肯定句，any用于否定句和疑问句。例如： She wants some chalk. She doesn’t want any chalk. Here are some beautiful flowers for you. Here aren’t any beautiful flowers. 二、any可与not以外其他有否定含义的词连用，表达否定概念。例如： He never had any regular schooling. In no case should any such idea be allowed to spread unchecked. The young accountant seldom (rarely, hardly, scarcely) makes any error in his books. I can answer your questions without any hesitation. 三、any可以用于表达疑问概念的条件句中。例如： If you are looking for any stamps, you can find them in my drawer. If there are any good apples in the shop, bring me two pounds of them. If you have any trouble, please let me know. 四、在下列场合，some也可用于疑问句。 1、说话人认为对方的答复将是肯定的。例如： Are you expecting some visitors this afternoon?（说话人认为下午有人要求，所以用some）Are you expecting any visitors this afternoon?(说话人不知道下午是否有人来，所以用any) Didn’t you give him some tickets?(说话人认为票已经给他了。问题的回答是：Sure I did) Did you give him any tickets?（说话人不知道是否给票了。回答可能是yes或no。） 2、款待用语或问句的实质等于提出一个要求并希望得到肯定的回答时。例如：

初中英语名词用法讲解

一、名词的分类 名词可分为普通名词和专有名词两大类。 1. 普通名词又可分为： （1）个体名词。如：cup，desk，student等。一般可数，有单复数形式。 （2）集体名词。如：class，team，family等。一般可数，有单复数形式。 （3）物质名词。如：rice，water，cotton等。一般不可数，没有单复数之分。 （4）抽象名词。如：love，work，life等。一般不可数，没有单复数之分。 2. 专有名词：如：China，Newton，London等。 二、名词的数 （一）可数名词的复数形式的构成规则 1. 一般情况下在名词的词尾加s，如：book books，pencil pencils. 2. 以-s，-x，-ch，-sh结尾的名词加-es，其读音为［iz］。如：bus buses，box boxes，watch watches，dish dishes等。 3. 以-y结尾的名词： （1）以“辅音字母+y”结尾的名词，把y改为i再加es，读音为［iz］，如：factory factories，company companies等。 （2）以“元音字母+y”结尾的名词，或专有名词以y结尾，直接在词尾加-s，读音为［z］。如：key keys，Henry Henrys等。 4. 以-f和-fe结尾的名词： （1）变-f或-fe为v再加-es，读音为［vz］。如：thief thieves，wife wives，half halves等。 （2）直接在词尾加-s，如：roof roofs，gulf gulfs，chief chiefs，proof proofs等。 （3）两者均可。如：handkerchief handkerchiefs或handkerchieves. 5. 以-o结尾的名词：