研究设计及撰写报告的一些建议

							研究設計及撰寫報告的一些建議


             黃熾森
     香港中文大學管理學系教授
地址: 香港新界沙田香港中文大學管理學系
 電郵:cswong@baf.msmail.cuhk.edu.hk
           2006年3月
                   大綱
(一) 恰當的資料提供者(Appropriate Informants)；
(二) 非隨機抽樣(Non-random Sample)；
(三)「沒回應偏差」(Non-response Bias)；
(四)「共同方法變異量」(Common Method Variance)；
(五) 多樣本的相互印證(Multi-Sample Cross-Validation)；
(六) 合併資料(Aggregating Data to Groups)；
(七) 資料的獨立性；
(八) 跨文化議題(Cross-Cultural Issues)；
(九) 研究方法議題的論文(Papers on Research Method
     Issues)；
(十) 跟從標準的報告方式。
    恰當的資料提供者-1
必須確保提供資料的人是有能力提供可信和有效的
 相關資料。
個人層面方面，關於個人的性格和態度的構念，由
 受訪者直接提供可能是最恰當的，但關於工作績
 效，其直屬主管則更為恰當。
關於機構層面的資料，一個普通的職員不一定了解
 機構的人力資源政策或機構的業績。
如果我們用郵寄方式要求機構填寫這些資料，便不
 能沒有其他的措施來確保資料提供者是否恰當的
 人選。
             恰當的資料提供者-2
…..a professor in Xiamen University helped us to distribute the survey
  to human resources and top-level managers in TNCs in Fujian, PRC,
  which constituted the final sample of this study….. we had good
  connections with the professor in Xiamen, Fujian, who could help us
  to collect quality localization data using his graduate
  students…..With the help of this professor in Xiamen, we sent out
  180 questionnaires to current and graduated MBA students who are
  top or middle-level managers in TNCs in Fujian Province. These
  managers were asked to fill out the questionnaires themselves if
  they were the human resources manager of the company. They
  were asked to refer to their human resources manager for
  necessary information if they were top executives of the company.
  After distributing the questionnaires and one round of telephone
  follow-up, a total of 139 responses were received. The response
  rate was 77%. The key advantage of this sampling process is the
  personal relationships between the Xiamen professor and the
  respondents. Thus, we were sure that the appropriate executives
  completed the questionnaires instead of other employees who may
  not have sufficient information to answer the questions.
        非隨機抽樣-1
在絕大部分的情況下，組織行為及人力資源
 管理的研究都無法做到完全的隨機抽樣
應該盡量避免太「隨意」的抽樣。
應盡可能有一個抽樣的框架(Sampling frame)
 或抽樣的標準。
提供樣本與研究問題可能相關的資料，讓讀
 者判斷樣本是否有偏差，在討論研究結果
 的部分，也應交待此一限制及可能對研究
 外部效度的影響。
                     非隨機抽樣-2
Final participants in our validation sample were 139 human resources
   managers of TNCs operating in Fujian Province in the PRC. We
   chose TNCs from one single province in order to control for the
   differences in governmental regulations because the policies
   towards foreign investments vary among local governments in the
   PRC. Fujian was chosen as the province for our sample for two
   reasons. First, Fujian is one of the most developed provinces in
   Southern China because of it geographic location. There are a wide
   variety of TNCs in Fujian from various countries such as Taiwan, the
   U.S., and other parts of the world….The parent companies of these
   operations came from various parts of the world, including Taiwan
   (25.8%), Hong Kong (17.2%), European countries (18.8%), the
   United States (15.6%), South Korea (10.1%), Japan (1.6%), and
   others (10.9%). The majority of these operations were in the
   manufacturing industries (56.9%). There were 12.3% in the financial
   industry, 4.6% in transportation, 3.9% in wholesale and retailing, and
   the remaining 22.3% in other industries. The majority of them had
   been operating in Fujian for at least three years (94.9%). The
   average number of years of operation of these companies in Fujian
   is 7.41 with a standard deviation of 3.49.
沒回應偏差(Non-response Bias)-1
與抽樣有關、又常常困擾研究人員的是回應率
 (response rate)偏低的問題，以不寄名而又不是
 即場回收的話，一般回應率都祗能有百分之二十
 幾左右。
可盡量收集資料以了解研究的效度是否會因較低的
 回應率而受到嚴重的影響，例如集中精力追蹤一
 小部分沒有回應者，比較他們與回應了的樣本是
 否有重大的分別。
如果我們有抽樣框架(Sampling frame)的資料，也
 可比較回應的樣本是否與整體框架內的情況有重
 要的分別。
沒回應偏差(Non-response Bias)-2
Although the average response rate for each survey is 49.3%, the final
   sample size that was used to examine the job perception-job
   satisfaction relationship was only 196 due to attrition and participant
   job changes. We therefore conducted three preliminary analyses to
   check the representativeness of this final sample. First, although
   the research participants were quite homogeneous in their
   backgrounds, they may have differed in three aspects: gender,
   undergraduate major, and year of graduation. The proportions of
   the respondents in these three aspects for the three surveys are all
   very similar to the general population. For example, the percentage
   of males in the population, and in the three surveys are 51.0%,
   52.0%, 49.5%, and 49.9%, respectively. Chi-square tests on these
   three aspects indicate no statistical differences (p>.10) for the three
   surveys. Second, we compared the level of job perception and job
   satisfaction in the first and second surveys for those who continued
   to respond in the next survey with those who did not. No significant
   differences are observed (p>.10). Thus, research participants who
   responded to all the three surveys did not appear to be a biased
   sample.
        共同方法變異量-1
(Common Method Variance)



                     X

        M (方法)

                     Y
      共同方法變異量-2
在組織行為及人力資源管理的研究中，由於很多時
  牽涉一些態度和行為的構念，所以這個問題會出
  現。
我們一定要處理，否則研究的內部效度會很受懷疑。
  我們可嘗試：
(1)以因子分析法，檢查來自同一方法的各測量項目
  背後共同因子的影響；
(2)同時測量一些無關或特別設定的變項(如Negative
  Affectivity)，然後把它與X和Y的共變量先行排除
  (常用的是迴歸方法)，最後才檢定X和Y的關係。
                 共同方法變異量-3
…..measures of this study are self-reported and so the effect of
  common method variances (Spector, 1992, 1994) or negative
  affectivity (Burke, Brief & George, 1993) may have affected the
  results. To deal with this potential threat, we have utilized the
  longitudinal design to examine the one-year, cross-lagged effects
  between job perception and job satisfaction. This design could help
  reduce the common method variances among the measures
  (Spector, 1992; 1994)……Nevertheless, we conducted the same
  LISREL analyses after partialling out some irrelevant variables,
  which should not have a direct relationship with both job perception
  and job satisfaction. In designing the study, we have included two
  irrelevant measures. The first one is a five-item measure (average
  coefficient alpha for the three surveys is .76) of need for achievement
  (Steers & Braunstein, 1976). The second one is a single item that
  requires the research participants to respond on a five point Likert-
  type scale (ranging from strongly agree to strongly disagree) to the
  statement: “I am an average worker”. After partialling out each of
  these irrelevant measures, the LISREL results are similar except that
  the path coefficients became smaller. Thus, common method
  variances did not appear to invalidate the results of this study.
              共同方法變異量-4
In addition to the four subjective questions, we added an
   objective indicator of localization success. Specifically,
   we used a ratio of the “Number of local managers
   occupying positions originally occupied by expatriates” to
   the “Total number of positions occupied by expatriates
   when the PRC operations started.” The numerator is a
   measure of actual localization success, while the
   denominator is a comparison base of the starting
   number of expatriate positions. This variable is very
   important because it allows us to double check the
   validity of the subjective indicator of localization success.
   Also, unless the respondents deliberately lied to us, this
   objective indicator can be a good dependent variable
   that has little respondents‟ biases with the independent
   variables.
         多樣本的相互印證
(Multi-Sample Cross-Validation)-1


「抽樣誤差」(Sampling error)是很普遍的，
 因此，要增加研究的效度和提高其品質，
 在可能的情況下，應考慮用不同的樣本來
 互相印證研究的結果。
              多樣本的相互印證-2
To investigate the distinctiveness and potential utility of the EI construct,
   we used a two-study-four-sample design with participants from Hong
   Kong and the People‟s Republic of China (PRC). In Study One, we
   used confirmatory factor analyses to show that, when properly
   defined and measured, EI is related to yet distinct from personality
   dimensions and has incremental predictive power on life satisfaction.
   In Study Two, Multi-Trait Multi-Method (MTMM) analyses were
   utilized to examine the construct validity of self-reports and others‟
   ratings of EI. In a student sample, parents‟ ratings explained
   additional variance in the students‟ life satisfaction and feelings of
   powerlessness after controlling for the Big Five personality
   dimensions. The results of the MTMM analyses were cross-
   validated on an employee sample with peer ratings of EI. These
   peer ratings were found to be significant predictors of job
   performance ratings provided by supervisors, after controlling for the
   Big Five personality dimensions.
              多樣本的相互印證-3

We have collected data from two samples in this study. The first sample
  is a pilot study designed mainly for item generation. In drafting the
  pilot-testing questionnaire for the first sample, we held several
  meetings with four practicing human resource managers….. generate
  measurement items for these studies…..All these four human resource
  managers worked for companies that have operations in the PRC.
  After the survey questionnaires were finalized, they were sent out to
  members of the Employers‟ Federation of Hong Kong and the
  American Chamber of Commerce……Forty-six member companies of
  the association who had operations in PRC completed and returned
  the pilot test questionnaire. The parent companies of these 46
  respondents came from Hong Kong (16), U.S.A. (12), U.K. (7), Japan
  (5), France (1), Germany (1), Switzerland (1), and Denmark (1).
           合併資料
(Aggregating Data to Groups)-1
雖然HLM可以處理跨層次的構念，但是，它的依變項一定要是最低層次
 的，否則不能應用。

小組的研究:
 Y=小組的績效
 X1=機構對小組的重視程度
 X2=小組成員的能力
由於Y是在第二個層次的，不能用HLM來處理。

工作設計的研究:
 Y=每一類工作的意外率
 X1=工作的性質(複雜性)
 X2=員工的因素(安全知識的水平)
由於Y是第二個層次的構念，不能用HLM來處理這個的研究問題。
       合併資料-2
把第一個層次的資料合併成第二個層次:
(1) 取小組所有成員能力的平均得分;
(2) 取從事同一類工作的員工安全知識的平均
    得分。
之後可完全在第二個層次中進行分析。
如果這樣做，首先要確定這些平均後的得分
    能有足夠的變異量，否則繼續用來分析便
    沒有意義了。
                  合併資料-3
Bartko, J.J. (1976) 提出的Intraclass Correlation:
先把第一層次的變項作ANOVA分析(例如以小組為單位，對
能力得分進行ANOVA分析；以工作種類為單位，對安全知
識得分進行ANOVA分析，取得MSB 和MSW 的結果後，用
以下的方程式計算(k是組別數目)：

                MSB  MSW
    ICC (1) 
              MSB  ( k  1) MSW

這個ICC(1)係數代表了組別間的變異量佔這個變項的總變
異量的比率，因此它愈大，代表把每一組的數據合併的問
題不大。
                           合併資料-4
Two reasons may explain the low intraclass correlation for the motivational
  value of the job. First, the intraclass correlation correlation ignored all
  single-respondent jobs in its calculation. The amount of between job
  variance for the group with multiple respondents was less because
  there were more low level jobs. Second, the intraclass correlation
  depended on the number of respondents per job.…..Two analyses were
  conducted to further investigate the within job respondents‟ agreement
  on the motivational value of the job. First, the intraclass correlation for
  the 23 jobs with more than two respondents was calculated. It was .40,
  similar to a recent study when the same analysis was conducted for a
  sample with about five respondents per job. Second, the two analyst
  ratings of the single-respondent jobs were added to the sample as if
  they were from two additional incumbents of each job. This allowed the
  high level single-respondent jobs to be included in the calculation. The
  intraclass correlation was .43, despite the average number of
  “respondents‟ of only 3.4. When the analyst ratings of the single- and
  two-respondent jobs were added to the sample, the intraclass
  correlation was .52 with an average number of respondents equaled to
  4.1. From these analyses, it appeared that there could be reasonable
  agreement between incumbents on the motivational value of their jobs,
  if a large enough number of incumbents existed or could be
  approximated.
                       合併資料-5
其實在HLM的分析中，也可應用ICC(1)來證明第二
 個層次的分析是有意義的。例如Liao and Chuang
 (2004)也有計算這個係數：
Null model. Our hypothesis predict that both individual- and store-level
  variables would be significantly related to employee service
  performance. In order for these hypotheses to be supported, there
  had to be significant between-store variance in employee service
  performance. Thus, using HLM, we estimated a null model in which
  no predictors were specified for either the level 1 or level 2 function
  to test the significance level of the level 2 residual variance of the
  intercept (τ00=.35, p<.001). The ICC(1) was .12, indicating 12
  percent of the variance in employee service performance resided
  between stores, and 88 percent of the variance resided within stores.
        資料的獨立性 -1
一個常見的問題，那就是當牽涉由主管提供部屬的
 工作績效時，往往是同一個主管評核數名部屬，
 如果我們的分析是以部屬為單位，例如每名部屬
 提供自變項(如工作態度；X)，而主管提供關於部
 屬的依變項(如工作績效；Y)。
問題是，如果這些主管之間對工作績效有很不同的
 看法，例如一些很寬鬆、有些很嚴格，這樣，由
 於依變項(Y)的不獨立性，是否會影響研究結果的
 效度呢？
計算James, Demaree and Wolf (1984)提出的
 Interrater reliability (RWG)係數。
          資料的獨立性 -2
如果在同一組別中，對某一變項(X)有幾個測量的得分結果(例如x1, x2,
  x3, x4, x5)，這五個得分是否相關？相關的程度又有多大？

RWG是要比較兩個變異量:
(1) 當這幾個得分完全無關時它們應有的變異量(σEU2)
(2) 收集到的數據中我們觀察所得的變異量(Sx2) ：
          RWG = 1 – (Sx2 / σEU2)

如果RWG等於1.00(即Sx2等於零)，這幾個得分百分之百吻合和相關；
如果RWG等於零(即Sx2等於σEU2)，這幾個得分是完全無關，是獨立的。
         資料的獨立性 -3
一般來說，RWG是當我們要求不同的人(rater)來評估
 同一個特質，然後從他們評分的相關情形來判斷
 評分是否可信，因此稱為「Interrater reliability」，
 所以係數愈接近1.00愈好。
但是，當我們應用RWG在檢定同一主管對不同部屬
 的工作績效評分的獨立性時，我們希望看到的是
 較小的RWG ，因為這樣證明了這個主管對某一部
 屬的工作績效評分並不受其他部屬的評分影響。
 所以，愈多主管的RWG較小，便代表這個變項的
 評分愈獨立。
                  資料的獨立性 -4
We conducted another preliminary analysis for the performance data
  because in our data, 41 supervisors rated the performance of more
  than one subordinate. Independence of the performance data may
  have created a problem in data analysis. Thus, we calculated the
  within-group inter-rater reliability for these 41 supervisors according
  to the formula provided by James, Demaree and Wolf (1984). To be
  conservative, we did not consider any response bias and assumed a
  triangular null distribution. The mean inter-rater reliability for the 41
  groups of performance ratings was 0.65 and its standard deviation
  was 0.31. Over half of these reliability coefficients (53.7%) were less
  than 0.70. George and Bettenhausen (1990) argued that an inter-
  rater reliability greater than 0.70 could be considered as an indicator
  of good within group agreement. From this result, we believe that
  the performance ratings may be regarded as independent and the
  results will not be affected significantly.
跨文化議題(Cross-Cultural Issues)-1
由於組織行為及人力資源管理的科學研究起源於西方社會，
 而以北美(尤其是美國)為主導，很多重要的學術期刊也以
 美國為基地，因此在非西方的社會進行研究，很多時便會
 被視為是「跨文化」性質的研究。這當然不一定正確，但
 是我們確實可能會面對這樣的誤解及從而引起的困難。
真正跨文化的研究，自然應有一定的要求，例如文化的定義
 是否合理，不同文化背景的樣本是否在其他非文化的特質
 上沒有差異等等。但在這裡不是要討論如何進行跨文化的
 研究，應另有專書討論。
在非西方社會中進行組織行為及人力資源管理的科學研究時，
 如何可以避免別人的誤解，及更公平地看到我們的研究對
 科學知識的貢獻。
           跨文化議題-2
研究沒有文化限制的普遍構念：
在對情緒智能(Emotional Intelligence)的研究時，用
 的都是中國人的樣本，結果有一位評審質疑研究
 結果是否在其他文化環境中也是正確的。
我們的觀點是：在我們定義之下的情緒智能構念(即
 人處理情緒的能力)是一個人類普遍的構念，我們
 的理論也沒有牽涉文化的議題，所以並不同意這
 個評審的意見，結果把他和編輯說服了。
                              跨文化議題-3
Our second limitation is that all the data in this project were collected in Hong Kong and the
   PRC. Cross-cultural generalizability of the results may be a concern. We do not know
   whether EI would vary across different cultures. However, when we go back to the EI
   literature, we do not find any discussion of EI across cultural boundaries. Our position is
   that one‟s ability to understand, regulate, and utilize one‟s emotions in constructive ways
   are general human abilities. There is no immediate evidence that the validity of EI, as
   defined under our four-dimensional view, should vary across culture. While further
   studies may be needed to verify this position, we take the general scientific
  attitude that psychological and management phenomena are considered
  as universal unless there are theories or evidence showing their cross-
  cultural variations.
While the EI construct may be universal, we agree that behaviors resulting
  from the EI of an individual may vary across culture. For example, a non-
   reactive quiet response by the subordinate when one‟s boss is making unreasonable
   demands may reflect high EI among Chinese but probably not among non-Chinese. In
   this respect, our use of self-report measure of EI may, in fact, be a plus because we
   asked respondents about their final judgment of the EI of the target person irrespective of
   the assessment clues or methods they would use. By doing so, we may be able to avoid
   some cross-cultural differences in expressing emotions or diagnosing emotions because
   the assessors would be able to use the clues or methods that are appropriate for their
   specific culture. This issue may have to be considered when behavior- or outcome-
   oriented tests of EI, such as the MSCEIT, are used across cultural boundaries.
           跨文化議題-4
發展或應用本土的構念：
在這類研究中，的確是建基於某一文化背景來發展
    一些新的構念，例如中國的「關係」(Guanxi)概
    念，但是，要在科學的領域中有所貢獻，這類研
    究最少要做到兩點。
(1) 首先是合理地論證無論是在理論上、或實證上，
    它都與現存西方或其他地方發展而已在文獻中報
    告的構念不一樣；
(2) 第二點是在實證工作時提出在現存構念之上的
    「增加效度」(incremental validity)，這樣我們
    提出新的構念也可能變成一個適用於其他文化背
    景的構念了。
跨文化議題-5
關係構念的例子
        跨文化議題-6
中國議題：

清楚地說明我們關心的原來便是中國的議題，
 所以我們可以參考西方或其他地方發展的
 理論和知識，但必須從中國的情況出發，
 例如關於外資企業在中國起用本地的中高
 層管理和技術人才的理論模型，便不太可
 能與其他國家完全一樣。
      跨文化議題-7
西方或其他理論在中國的適用性：

研究問題是驗證西方或其他地方發展的理論是否同
 樣適用於解釋中國社會的現象，當然，我們不應
 該盲目驗證，最好是基於現有理論的來龍去脈和
 對中國社會現象的分析，首先提出清晰的假說，
 例如這些理論應如何修訂，才可更準確地描述中
 國社會的現象，最後才進行驗證這些假說的針對
 性研究。
                     跨文化議題-8
In contrast to the educational system in the US, high school students in
    Hong Kong are assigned to science or arts streams after Grade 9.
    Except in a small number of special schools, most students are
    strictly confined to subjects within their stream of study. For
    example, a science student can study only subjects such as physics
    or chemistry and is not allowed to study literature and history.
    Similarly, a student who is assigned to the arts stream is generally
    not allowed to take chemistry no matter how strong his/her interest
    is in the subject. As a result, students in Hong Kong are basically
    separated into two groups (science vs arts) and confined to subjects
    within their groups after Grade 9. This education system is a good
    example of the environmental variable discussed by Holland (1985:
    15-18) that would affect the students‟ formulation of vocational
    orientations.
                      跨文化議題-9
The second characteristic of the education system in Hong Kong is its
  strong bias towards the science stream. Due to the design of the
  examination syllabus and teaching methods, it has been a long
  tradition in Hong Kong that students have a strong positive bias
  towards the science stream in their high school education. The arts
  stream is widely perceived by students as the course of study for
  those who are not smart enough or strong enough to deal with
  mathematical and scientific symbols. As a result, students in the
  science stream are usually academically brighter and quite
  homogeneous in their abilities and interests, while students in the
  arts are much more varied. These special environmental
  characteristics lead us to the hypothesis that science students are
  supposed to have an interest in realistic and investigative jobs while
  arts students will have interests in other jobs (i.e., artistic, social,
  enterprising, and conventional jobs) in Hong Kong.
                跨文化議題-10
A final characteristic of the Hong Kong environment is that
   Hong Kong is a society with a strong business culture
   and is strongly influenced by Chinese traditions. As a
   result, interpersonal relationships or guanxi play very
   important roles in the business world (Alston, 1989;
   Hwang, 1987; Law et al., 2000). Davies (1995) reported
   that 98 per cent of the 150 Hong Kong Chinese
   executives who responded to their survey confirmed that
   „personal connections with local Chinese organizations‟
   form an extremely important part of their business lives.
   As a result, we hypothesize that the enterprising and
   social types would be strongly and closely related to
   each other and would be even stronger than the R-I, E-C
   and S-A links.
                     跨文化議題-11
Figure 1. The two-group model as compared with Holland’s model.



           R             I                             R        2          I
                 2                                             5 5
       1                                          3    6                   6       3
                 3                                                     7
                                                               7
 C                                 A C                4
                                                           8
                                                                               4
                                                                                       A
                                                  2                                2


           E             S                             E           1
                                                                           S
           Holland’s model                         The revised model in Hong Kong


       Note: Lines with the same width and shape mean that the correlations between the

two sets of personality dimensions are the same; the smaller the number, the greater is the

correlation.
   研究方法議題的論文-1
一些牽涉研究方法議題的研究，很多時人們
 都以為這類議題一定是計量的，以數學統
 計為主，但並不一定如此。

似乎最重要的還是提出的「研究問題」是否
 恰當和重要，具體的數學統計推論是次要
 的。
     研究方法議題的論文-2
構念的定義及量化的問題：

「結構方程模型」(Structural Equation Model；
 SEM) 是以因子分析法來估計「潛在構念」的，
 但是，並不是所有「多構面構念」
 (Multidimensional Construct)都可以這樣估計的，
 例如在組織行為及人力資源管理學中常用的工作
 性質和工作滿足感便不是這樣的。但是，當時應
 用SEM的論文似乎都沒有注意這一點。
       研究方法議題的論文-3
Latent Model: constructs as underlying
  higher-order abstractions behind their
  dimensions, e.g., g-factor
         研究方法議題的論文-4
Aggregate Model: constructs exist at the same level as
  their dimensions and are formed as a mathematical
  function of their dimensions, e.g., job satisfaction, job
  perception
         研究方法議題的論文-5
Profile Model: constructs exist at the same level as their
  dimensions and are formed as various combinations of
  their dimensional characteristics, e.g., personality




        High/Low         High/Low         High/Low
        研究方法議題的論文-6
…..leads to different conclusions about the relative
 effects of job perception and social interactions
 on job satisfaction. Under the factor (latent)
 model, both job perception and liking are
 significant predictors of job satisfaction. In
 contrast, only job perception is a significant
 predictor of job satisfaction under the composite
 (aggregate) model. In other words, liking would
 be concluded as an important antecedent of job
 satisfaction under the factor model, but not
 under the composite model.
     研究方法議題的論文-7
SEM可以用
 「nonrecursive
 model」來驗證兩個
 構念互為因果的關
 係，雖然在數學方
 法上可行，但它違
 反了基本的邏輯：
 兩個構念怎可能在
 同一時間點上互為
 因果呢？
        研究方法議題的論文-8
Hunter & Gerbing (1982):
 “the problem that nonrecursive causation pose
 in cross-sectional models can be seen by
 considering the implications that a two-way
 arrow has for indirect effects. If X and Y have an
 effect on each other, then X has an impact on Y,
 which has an impact on X, which has an impact
 on Y…this cycling can go on (for) as many steps
 as can be imagined. In reality, there is no such
 instantaneous cycling process. (pp.288-289)”
研究方法議題的論文-9
       研究方法議題的論文-10
其中一個模疑結果清楚顯示以nonrecursive model來驗證互
 為因果的關係，當樣本數是300時，我們有百分之四十七
 的機會作出錯誤的結論：

「In the second example, X and Y in the true time-lagged
 model were set to be reciprocally related. The path
 coefficients from both X to Y and Y to X were set to
 be .25…..Finally, M3 (the nonrecursive model)….. gave
 very poor estimates of the true population
 parameters…..the point estimates of the path from X to
 Y and vice versa were .18 and .34. The probability of
 concluding a reciprocal relation between X and Y was
 only 53% when N was 300.」
   跟從標準的報告方式-1
在進行組織行為及人力資源管理的科學研究和撰寫
 報告時，盡量不要標奇立異，應以踏實和嚴謹為
 主要考慮。

學術研究的工作，在於提出足夠的證據以支持我們
 的論點和結論，我們應力求以最簡潔易明的方式
 陳述論點及報告證據，然後作出結論。

應盡量把複雜的概念和論據以簡單的方法說明，而
 不是把簡單的現象用艱深專門的術語來表達。
    跟從標準的報告方式-2
科學研究的工作是有內在標準的:
(1)用合符邏輯理性的方式建構理論和假說以描述現象
(2)用客觀系統的方法搜集資料來驗證理論和假說。

第一部分一定是說明「研究問題」，要說服讀者為什麼這是
 重要的及尚未有答案、或最少未有定論的問題，然後對構
 念和假說加以說明；
第二部分要報告研究採取的方法，最少要包括抽樣過程、測
 量及分析方法；
第三部分是報告分析資料的結果，一般而言先要報告各變項
 的平均、標準差、信度及相關係數，然後是驗證假說所採
 用的分析結果，這部分通常會以圖表來輔助說明；
最後一部分是討論整個研究的結果，一般應包括對「研究問
 題」的結論、研究的限制及對未來研究的建議。
這是基本的功夫，應盡量習慣用這樣的方式來作報告。
   跟從標準的報告方式-3
嚴耕望先生：

「我從事學術工作五、六十年，發表的論文
 很少受人質疑，為什麼呢？因為我有十分
 證據，也祗講八分話，不單不說過頭，還
 一定留有餘地。」