Data Analysis

Sampling

We should reduce MoE=1.96*開方p(1-p)/n

We don't have to collect same no. of participants between population groups because we have a technique called weighting

Compare the distribution of your interviewee population proportion with the official proportion.

Data Cleaning (before analysis)

"wrong" data: missing data, out of target participants, logic impossible, multiple and for single-answer questions, inconsistent answers for multiple questions

Solution
1) Delete all records of that respondent (if the contaminated part is too big);
2) Delete the contaminated part only;
3) Adjust the answer and/or infer the real value (未教)

Step 1

Descriptive Statics:
1) Central Tendency: Mean, Median, and Mode (everybody is mode when there is no duplicate)
2) S.D.
3)

Step 2

Types of Scale:
1) Nominal: Have no hierarchy (Gender, Ethnicity)
2) Ordinal: Difference between values is no the same (e.g. ranking: the difference between 1st and 2nd as well the difference between 2nd and 3rd)
3) Interval: have no absolute zero and Difference between values is the same
4) Scale: have absolute zero and Difference between values is the same

Step 3

Hypothesis Testing

Step 1: Set the null hypothesis (=there's no difference): H0.
Calculate the probability (p-value)

Results:
1) p-value smaller that 0.05 and we reject the H0: Correct Rejection OR Type I Error
2) p-value larger that 0.05 and we accept HO: Correct Acceptance OR Type I Error

Step 2: p-value>0.05 accept; p-value<0.05 reject

Step 4

Chi-squared Test? T-Test?

Step 5

Correlation

cannot imply causation

While A and B show correlation: We don't know A or B is the cause or result; May be C is the cause of A and B

The fallacy of interpretation

You may use box plots/scatter plot to identify outliner which may be wrong.

Historical Archive Analysis

The Pursuit of History: Aims, methods and new directions in the study of modern history (John Tosh and Seán Lang, 2006)

Using Source

External Criticism

2) Is the content consistent with known facts?

1) Who wrote it? The origin of the source.

3) Thirdly, the form of the document may yield vital clues. The historian who deals mostly in handwritten documents needs to be something of a palaeographer in order to decide whether the script is right for the period and place specified, and something of a philologist to evaluate the style and language of a suspect text.

4) Chemical testing

Internal Criticism

The interpretation of the content

1) Word meaning may different time to time. Never "reading modern meanings into the past"; be aware of "different levels of meaning may have been embedded in the same text" (P. 96)

2) Reliability

2a) Does he actually present? Learned second hand? (P. 96-97)

2b) Put to pen immediately? Put to pen after the sharpness of his/her memory had blurred? (P. 96-97)

2c) the intention and prejudices of the writer (P. 98)

"Autobiographies are notorious for their errors of recall and their special pleading." (P. 99)

"Even the most tainted sources can assist in the reconstruction of the past." (P. 100)

Reading Source

1) Social context of specific archive

Public Records

Social Context: that may be unreliable as there are bias or being censored

Officially published records

"prefer to go to the originals". "If these are not available, the published versions must be scrutinized carefully, and as much as possible must be found out from other sources about the circumstances in which they were compiled." (p. 103)

Principles

1) "amass as many pieces of evidence as poss- ible from a wide range of sources – preferably from all the sources that have a bearing on the problem in hand" (P. 103)

2) "Each type of source possesses certain strengths and weaknesses; considered together, and compared one against the other, there is at least a chance that they will reveal the true facts – or something very close to them." (P. 103)

3) "The need for primary evidence from ‘insiders’ as well as ‘outsiders’ is an important guideline for historical research, with wide ramifications." (P. 104)

"a golden rule that both sides of a diplomatic conversation must be studied" (P. 104)

Political Science in History: The public records tend to give too much prominence to administrative considerations (thus reflecting the principal interest of the civil servants who wrote most of them) and to reveal much less about the political pressures to which ministers responded; hence the importance of extending the search to the press and Hansard, private letters and diaries, political memoirs, and – for recent history – to first-hand oral evidence.

Unwitting evidence

Working backwards from the sources: While a culture was only depicted by an unreliable author, you have to visit that community to understand that culture. However, as the time passed, what you see is not what those author see.

4) "No document, however authoritative, is beyond question" (P. 109)

"Divergent sources have to be weighed against each other, forgeries and gaps explained." (P. 109)

*"the exact procedures vary according to the type of evidence" (P. 110)