CH 16 Data Handling & Preparation/ Data Handing and Data Coding Lecture 9

steps in data preparation for analysis

• coding

• editing, validating and cleaning

• entering

• transforming

1. data editing, 2. coding, 3. statistically adjusting data if required

data editing

to identify omissions, ambiguities, errors, conducted in field by interviewer and field supervisor

problems to identify in editing

• interviewer error ( may not be giving respondent correct instructions)

• omissions (fail to answer all questions)

• ambiguity (response unclear or not legible)

• inconsistencies (2 respondes inconsistent)

• lack of cooperation (rebeling & checking same response)

• ineligible respondent (inappropriate respondent (ex. women only 18)

alternatives available for data editing

• contact respondent again done by interviewer

• throw whole questionnaire as not useable

• throw out problem questions

• bypass questions

• code illegible to idk or no opinion to simplify

• input missing values for certain variables through use of mean profiles

by product of editing

helps in evaluating and guiding the interviews; an interviews tendency to allow a certain type of error should be detected by editing

flawed and missing data in editing

• discard flawed records with missing answers

• treat flawed record as missing data and treat missing data as separate category

• obtain additional. information

data coding

• one or multiple columns for each question, each row corresponds to 1 response

• indicators or identification #s assigned for each type or responses

• different code for diff. answers and missing answers

• assign same code for similar answers

data coding open ended q

• specify exactly how responses to be entered

• column ref synonymous with variable identification link b/t question # columns for each response

• each q in separate column and range of permissible values provide key info of the value to be entered

• response value into spreadsheet to generate info but needs to be checked for error

• assign same code for similar answers

data coding close ended q

• long list of possible responses generated & placed into list of items

• assignment of a response involves a judgement decision if the response doesn’t match a list item exactly

• EX: q why did you select instrument from ajax. may get many responses, decisions must be made about response categories. difficult to put into categories

general rules

• one or multiple columns for each question, each row corresponds to 1 response

• indicators or identification #s assigned for each type of response

data coding example

which is your favorite online book store

1. amazon (A)

2. barnes and noble (B)

3. other (o)

records

#1 (2)

#2 (1)

#3 (3)

statistically adjusting the data

• to enhance its quality for data analysis

• weighting

weighting

• procedure by which each response in database is assigned a number according into some pre specified rule.

• done to make sample data more representative of a target population on specific characteristics

• underrepresented categories given higher weights

• over-representated given lower weights

• used for adjusting the sample so that greater importance is attached to respondents with certain characteristics

example: if a study is conducted to determine market potential of new sports drink → researcher might attach greater weight to opinion of young people because they main users of product

ex: 2.0 to person who under 30

1.0 to person who over 30

variable respecficiation

• a procedure in which the existing data are modified to create new variables or in which large number or variables are collapsed into fewer variables

• purpose: create variables consistent with study objectives

ex: original variable represented the reasons for purchasing a car w/o response categories. may be put into 4 categories: performance, price, appearance, service

dummy variables

• variable 0 or 1 to represent categories

• uses extensively for respecifying categorial variables

• also known as binary, dichotomous, instrumental and qualitative var.

• m levels of qualitative we use m-1 to specify them

• m categories→ m-1 dummy var

ex: 2 categories (1st half , 2nd half, m=2 m-1-1 so 1 dummy needed)

scale transformation

• for interval, ratio data

• manipulation of scale values to ensure comparability with others scales

• in same study different scales be employed for measuring different variables

• standardization

standardization

• type of scale transformation

• allows researchers to compare variables that have been measured using different types of scales

• ex: scales measures in $$, cents, value of variance you need to change

• to compare variances both variables bought down to common unit of measurement

Zi (Xi - X) / Sx

X has _ on it

strategy for data analysis

1st step in data analysis after data preparation is to analyze each question or measure itself by using tabulation

tabulation

consists of counting the number of cases that fall into various categories or counting how many responses fall into each category.

ex: whats your fav color tabulation into a table red:2, Orange: 1, green: 2

primary use for tabulation

1. determining empirical distribution (frequency distribution) of the variable in question (ex. 40% said blue, 10% said orange)

2. calculating the descriptive (summary) statistics particularly means or %’s (can compute % , totals helps spot missing/weird mistakes)

cross tabulations

• compares 2 variables at same time instead of 1

• to assess if any associations is present instead separate b/t two nominal and ordinal variables (non number)

• if variables measured as interval or ratio they can be transformed to nominally scaled variables

• sample is divided into subgroups in order to learn how the dependent variable varies from subgroup to subgroup

• % computed on each cell basis or rows or columns, when computations are by rows or columns cross tab usually referred as contingency tables

ex: the income of household can be rescaled as <30,000 and 39,000 to cross tab with another nominally scaled variable

Frequency Distribution

• simply reports the number of responses that each question received and its the simplest way of determining empirical distribution of the variable

• part of tabulation

• organizes data into classes or groups of values and shows the number of observations from the dataset that falls into each of the classes

• response categories may be combined with many question

• result in categories w worthwhile # of respondent

• histogram: series of rectangles each proportional in width to the range of values with a class proportions in weight to # of items falling into class

descriptive statistics

statistics normally associated with a frequency distribution that helps summarize information at frequency table

includes

• measures of central tendency (mean, median, mode)

• measures dispersion (ranges sd, coeff of variation)

• measures of shape (skewness & kurtosis)

requires data to be collected using interval or ratio scaled questions

factors influencing the choice of statistical techniques

1. type of data

2. research design

3. assumptions underlying the test statistic & related considerations

type of data

nominal, ordinal , interval, ratio

nominal

most primitive form of data, #s assigned to objects, based on objects, belong to particular categories made important

ordinal

represents a higher measurement than nominal, because numbers assigned to reflect order also to identify the objects mean, median and nonparametric tests

interval & ratio

metrics data best for data analysis, wide range of parametic, non parametric test mean, median, mode

research design

• a second consideration that affects the choice of analysis is RD to generate data

• decisions analyst have to face involve the dependency of observations the number of observations per project, # of groups being analyzed & control excersised over variable of interest

• independent or dependent samples?