Spatial and Survey Statistics

A survey proposal is a formal written document prepared before conducting a survey. It is a plan that describes what is to be studied, why it is to be studied, how, where, and when the survey will be carried out, and how much it will cost. It serves as a blueprint for the researcher and as a request for approval or funding from a supervisor, department, or sponsoring agency. A good proposal convinces the reader that the survey is feasible, well-designed, and worth carrying out.

Components of a survey proposal:

1. Title / Topic — a clear, specific statement of the survey problem.
2. Introduction and Background — brief context, review of related literature, and rationale (why the survey is needed).
3. Objectives — general and specific objectives the survey intends to achieve.
4. Hypotheses / Research Questions — tentative assumptions or questions to be tested/answered.
5. Scope and Limitations — the area, population, and boundaries covered, and what is excluded.
6. Methodology — survey design, population and sampling technique, sample size, data collection tools (questionnaire, interview, observation), and data analysis methods.
7. Data Collection Procedure — how, by whom, and in what sequence data will be collected; pretesting/pilot survey.
8. Time Schedule (Work Plan) — a Gantt chart or timetable of activities from preparation to report writing.
9. Budget / Cost Estimate — estimated expenditure on personnel, travel, printing, stationery, data processing, etc.
10. Expected Outcomes / Significance — the contribution the survey will make to knowledge or policy.
11. References — sources cited in the proposal.
12. Appendices — draft questionnaire, maps, tables, and other supporting material.

Thus a survey proposal is a complete, systematic plan covering the problem, objectives, methodology, schedule, and budget, which guides the entire survey work.

Structured and Semi-structured Instruments

Structured instrument: A structured (or standardized) instrument is a data collection tool in which both the questions and the possible responses are pre-determined and fixed in advance. The respondent has to choose from given alternatives, so there is little or no freedom to answer in his or her own words. Examples: questionnaire with closed-ended questions, rating scales, checklists, and interview schedules with fixed alternatives.

Semi-structured instrument: A semi-structured instrument is a tool in which some questions are pre-determined (closed-ended) while others are open-ended, allowing the respondent or interviewer to probe further and add details. It lies between the fully structured and the fully unstructured types. Examples: semi-structured interview guide, focus group discussion guide, and questionnaire containing both closed and open questions.

Characteristics

Characteristics of structured instruments:

1. Questions and response alternatives are fixed and pre-planned.
2. Mostly closed-ended; responses are easy to code, tabulate and quantify.
3. Highly reliable and objective, as every respondent is asked the same questions in the same order.
4. Easy and quick to administer, even to a large sample.
5. Less flexible; the respondent cannot express feelings, opinions or experiences beyond the given options.
6. Suitable for descriptive and survey-type research where the variables are already known.

Characteristics of semi-structured instruments:

1. A mix of closed-ended and open-ended questions.
2. Flexible; the interviewer may probe, clarify or add questions as needed.
3. Allows the respondent to express views, experiences and reasons in his or her own words.
4. Provides both quantitative (coded) and qualitative (narrative) data.
5. Requires a skilled interviewer and more time for administration and analysis.
6. Suitable for exploratory and explanatory research where the researcher wants depth along with some comparability.

Conclusion: Structured instruments give standardized, comparable and easily analyzable data but lack depth, whereas semi-structured instruments give flexibility and richer information but demand more skill and time. The choice between them depends on the objectives, nature of the study and the resources available.

Importance of Web-Based Questionnaires

A web-based questionnaire is a survey instrument hosted on a website or web platform that respondents fill in online, usually through a link sent by email or shared on social media. It has become important because it replaces the slow, costly paper-and-pen survey cycle with an automated one. Responses are captured directly in digital form, so no manual data entry is needed, and results can be tabulated in real time. It also allows a researcher to reach a geographically scattered population at very low marginal cost, which is essential for academic research, market studies, and organizational feedback where the sample is large or dispersed.

Advantages

1. Low cost — no printing, postage, or telephone charges; a single link reaches thousands of respondents.
2. Wide and fast reach — respondents from any location and time zone can participate, and data arrives within days rather than weeks.
3. Automatic data capture — answers are stored in a database, eliminating transcription errors and manual coding.
4. Real-time analysis — results can be monitored and analyzed while the survey is still running.
5. Flexible design — multimedia (images, audio, video), skip logic, and branching questions can be used, and questions can be randomized to reduce order bias.
6. Convenience and honesty — respondents answer at their own pace, and anonymity tends to increase willingness to report sensitive information.
7. Low interviewer bias — there is no interviewer to influence or pressure the respondent.

Disadvantages

1. Limited population coverage — only people with internet access and digital literacy can respond, which excludes rural, poor, and elderly groups and biases the sample.
2. Self-selection bias — those who choose to respond may differ systematically from those who do not, lowering representativeness.
3. Low response rate — online surveys are often ignored or abandoned midway, and there is little control over who actually fills the form.
4. No control over the environment — the researcher cannot verify who answered, whether the respondent understood the questions, or whether answers were given carelessly.
5. Technical problems — server downtime, broken links, browser incompatibility, and device differences can lose data or frustrate respondents.
6. Difficulty with open-ended questions — typing long answers is tedious, so qualitative depth is usually poorer than in interviews.
7. Security and privacy risk — personal data may be exposed to hacking or misuse if the platform is not secure.

In short, web-based questionnaires are important for their speed, reach, and cost efficiency, but they are best used when the target population is online and the questions are simple and structured.

Definition of Quantitative Data Source

A quantitative data source is any source that provides data in numerical or measurable form, which can be counted, quantified, and subjected to statistical analysis. Such data are expressed in terms of quantities, magnitudes, or frequencies (e.g., population size, income, rainfall, birth rate) and are used to draw conclusions through mathematical and statistical techniques. Examples include censuses, sample surveys, and vital registration records.

Can Administrative Data Be Considered a Quantitative Data Source?

Yes, administrative data can be considered a quantitative data source. Administrative data refers to information collected and maintained by government or other organisations primarily for their own operational, record-keeping, or service-delivery purposes — for example, birth and death registers, school enrolment records, tax records, hospital admission records, and migration registers. Since these records are maintained in numerical form (counts, totals, rates, categories with numerical codes), they yield quantitative information that can be tabulated and statistically analysed. Thus, administrative data qualifies as a quantitative data source, although it is a secondary source because it is not collected primarily for research or statistical purposes.

Purpose of Using Such Data

1. Low cost and quick availability — Administrative records already exist, so data can be obtained cheaply and without conducting a fresh field survey.
2. Complete coverage — Since records are maintained continuously for the whole administrative area, they often provide complete or near-complete coverage of the target population.
3. Time-series and trend analysis — Continuous recording allows comparison over time (e.g., yearly birth or death rates).
4. Planning and policy formulation — Governments use such data for development planning, resource allocation, and policy decisions.
5. Supplementary and cross-checking source — It supplements survey and census data and helps verify their accuracy.
6. Small-area and disaggregated data — It provides data for local levels, small areas, and specific groups where surveys may not reach.
7. Basis for indicators — It helps compute demographic and socio-economic indicators such as fertility, mortality, and literacy rates.

Resolution in Remote Sensing

Resolution refers to the ability of a remote sensing system to distinguish (separate) two closely spaced objects or to record fine details of a target. In other words, it is the smallest measurable or detectable difference that a sensor can record. A sensor with higher resolution can detect finer details, while a sensor with lower resolution records only coarse information. Resolution is a measure of the level of detail and accuracy of the data acquired by a sensor.

Types of Resolution

1. Spatial Resolution It refers to the smallest ground area (ground footprint) that a sensor can resolve as a separate object, usually expressed as the pixel size on the ground (e.g., 30 m, 10 m, 1 m). A smaller pixel size means finer spatial resolution. For example, Landsat has 30 m spatial resolution, while IKONOS has 1 m. It depends on the Instantaneous Field of View (IFOV) and the altitude of the sensor.

2. Spectral Resolution It refers to the ability of a sensor to distinguish fine differences in wavelength (electromagnetic radiation) and is described by the number, width, and position of spectral bands. A sensor with many narrow bands has high spectral resolution (e.g., hyperspectral sensors with hundreds of bands), while a sensor with few broad bands (e.g., Landsat MSS) has low spectral resolution.

3. Temporal Resolution It refers to the frequency with which a sensor revisits and images the same area of the Earth's surface, i.e., the repeat cycle or revisit time. For example, Landsat has a 16-day revisit period, while geostationary satellites provide near-continuous observation. High temporal resolution is useful for monitoring dynamic phenomena such as floods, crops, and clouds.

4. Radiometric Resolution It refers to the ability of a sensor to discriminate fine differences in the intensity (brightness) of the reflected or emitted radiation. It is expressed in terms of the number of quantization levels or bits — e.g., 8-bit data gives 256 (2⁸) grey levels, while 11-bit gives 2048 levels. More bits mean higher radiometric resolution and better detection of subtle tonal differences.

Conclusion Thus, resolution in remote sensing is the capability to detect detail, and it is of four main types — spatial, spectral, temporal, and radiometric — each describing a different dimension of the sensor's ability to record information about the Earth's surface.

Vector vs Raster Data Model

Similarities

1. Both represent spatial data — Each is a way of storing geographic features (location, shape and attributes) in a GIS/database.
2. Both use a coordinate system — Features are georeferenced to a common map projection and coordinate system.
3. Both store attributes — Non-spatial information is attached to spatial features in an attribute table.
4. Both support spatial analysis — Overlay, buffering, query and map production can be done with either model.
5. Both are digital models — Real-world objects are abstracted/simplified into a computer-readable form.

Differences

Conclusion

The vector model is suited to discrete, well-defined objects with accurate boundaries, while the raster model is suited to continuous phenomena varying over space. In practice, both are often integrated (e.g., raster DEM with vector road network) to combine their advantages.

Variable selection (also called feature selection) is the process of choosing the most relevant and useful variables (features) from a larger set of available variables to build a statistical or machine-learning model. The aim is to keep only those variables that contribute meaningfully to predicting or explaining the dependent (response) variable, and to discard irrelevant, redundant, or noisy ones. This improves model accuracy, reduces overfitting, shortens computation time, and makes the model easier to interpret.

How the process is done:

1. Start with all candidate variables collected from the data or domain knowledge.
2. Evaluate each variable (or subset) using a criterion such as correlation with the target, statistical significance (p-value, t-test, F-test), or a model performance measure (AIC, BIC, R², cross-validation error).
3. Rank or test variables according to that criterion.
4. Select the best subset — keep the variables that meet the threshold or give the best performance.
5. Validate the chosen subset on test data to confirm it generalizes well.

Ways (methods) of variable selection:

• Filter methods — variables are ranked by a statistical score independent of any model, e.g. correlation coefficient, chi-square test, ANOVA F-test, mutual information. Fast but ignores variable interactions.
• Wrapper methods — a model is trained repeatedly on different subsets and performance is compared, e.g. forward selection, backward elimination, stepwise selection, recursive feature elimination. Accurate but computationally expensive.
• Embedded methods — selection happens during model training itself, e.g. LASSO (L1 regularization), Ridge, Elastic Net, decision-tree/random-forest feature importance. Balances accuracy and cost.
• Dimensionality reduction (feature extraction) — original variables are transformed into a smaller set of new variables, e.g. Principal Component Analysis (PCA), Factor Analysis, Linear Discriminant Analysis (LDA).

In practice, filter methods are used first to remove obviously weak variables, then wrapper or embedded methods refine the final subset.

Major Problems in Survey Research

Survey research, though widely used, suffers from several practical and methodological problems:

1. Sampling problems — The sample frame may be incomplete or outdated, so the sample does not truly represent the population. Non-response bias arises when those who refuse or fail to respond differ systematically from respondents.
2. Respondent identification problem — It is often difficult to locate the correct, eligible respondent, especially in household or household-based surveys where the target person may be absent, or where "who actually answered" is unclear.
3. Response errors — Respondents may misunderstand questions, give socially desirable answers, deliberately falsify information, or suffer from recall bias (forgetting past events).
4. Questionnaire and interviewer problems — Poorly worded, leading, or ambiguous questions; interviewer bias, inconsistent probing, and variation between interviewers reduce reliability.
5. Coverage and accessibility problems — Difficult terrain, remote areas, illiteracy, language barriers, and lack of cooperation make some populations hard to reach.
6. Cost, time and administrative problems — Surveys are expensive, time-consuming, and require trained enumerators, supervision and data management.
7. Ethical and confidentiality problems — Invasion of privacy, risk of harm, and breach of confidentiality can affect both participation and data quality.

Reducing the Problem of Identifying the Respondent

The respondent identification problem can be reduced by:

• Clear definition of the target respondent — specify exactly who qualifies (age, sex, role, residence) before fieldwork.

2. Use of a proper sampling frame — updated voter lists, census lists, household registers, or institutional records to locate units accurately.
3. Household listing and Kish grid / random selection — list all eligible members in a household and select one randomly, so the choice is not left to the enumerator.
4. Unique identifiers — assign ID codes, addresses, GPS coordinates or phone numbers to each sampled unit to avoid duplication and substitution.
5. Repeated call-backs and appointment scheduling — visit at different times (evenings, holidays) to find absent respondents.
6. Trained and supervised enumerators — with strict instructions not to substitute respondents; verification by supervisors through spot checks and call-backs.
7. Pre-testing (pilot survey) — to detect identification difficulties before the main survey.
8. Use of mixed modes — telephone, online or proxy interviews where face-to-face identification is difficult (with proxy clearly recorded).

Advantages and Disadvantages of Survey Compared with Other Methods

Advantages over other methods

Disadvantages compared with other methods

Conclusion: Surveys are best for large-scale, descriptive, generalizable studies of opinions and behaviour, but for causal explanation, in-depth understanding or study of actual behaviour, experiments, observation and case studies are superior. Often a combination (triangulation) gives the most reliable results.

Remote Sensing

Remote sensing is the science and art of acquiring information about an object, area, or phenomenon without coming into physical contact with it. The information is obtained by measuring the energy reflected or emitted from the target using sensors mounted on platforms such as satellites, aircraft, or drones, and is later analysed to interpret the Earth's surface and environment. In the broad sense, remote sensing includes the entire process of data acquisition, processing, and interpretation.

Physical Basis of Remote Sensing

The physical basis of remote sensing rests on the interaction of electromagnetic (EM) radiation with matter. Every object on the Earth's surface has a characteristic spectral response — it reflects, absorbs, and emits EM energy in a manner determined by its physical and chemical properties. This variation in reflected or emitted energy is the fundamental source of information in remote sensing. The main elements of this physical basis are:

1. Electromagnetic radiation as the carrier of information. EM radiation travels through space as waves at the speed of light, characterised by wavelength (λ) and frequency (ν), related by c = λν. Remote sensing generally uses the visible, near-infrared, thermal infrared, and microwave regions of the spectrum.

2. Energy–matter interaction. When incident radiation falls on a target, three things happen: part is reflected, part is absorbed, and part is transmitted. By the conservation of energy, for any wavelength:

$\rho_\lambda + \alpha_\lambda + \tau_\lambda = 1$

where ρ = reflectance, α = absorptance, and τ = transmittance. Remote sensing sensors mainly record the reflected component (optical region) or the emitted component (thermal region).

3. Spectral signatures. The plot of reflectance versus wavelength for a given object is its spectral signature. For example, healthy vegetation reflects strongly in the near-infrared and absorbs in the red (chlorophyll absorption), while water absorbs most infrared energy. These distinctive signatures allow different land-cover types to be identified and separated.

4. Atmospheric effects. Radiation must pass through the atmosphere, where it is affected by scattering (Rayleigh, Mie, non-selective) and absorption by gases such as water vapour, CO₂, and ozone. Atmospheric windows — spectral regions of low absorption — are therefore chosen for remote sensing observations.

5. Radiation laws. The emitted radiation from a body is governed by Planck's law, Stefan–Boltzmann law, and Wien's displacement law, which relate temperature to emitted energy and peak wavelength. These laws allow thermal remote sensing to estimate surface temperature.

Operation of Energy Sources and Radiation Principles

The operation of remote sensing depends on a chain of energy transfer from source to sensor:

1. Energy source. The primary and ultimate source of energy is the Sun, which radiates energy according to Planck's blackbody law, with peak emission in the visible region (~0.5 μm). For passive remote sensing, sunlight serves as the source of illumination. For active remote sensing (radar, LiDAR), an artificial source such as a microwave or laser pulse is transmitted by the sensor itself, and the return signal is measured.

2. Propagation through the atmosphere. The emitted energy travels through the atmosphere and interacts with atmospheric constituents. Scattering and absorption modify the signal; hence sensors operate in atmospheric windows (e.g., visible, 8–14 μm thermal, and microwave windows).

3. Interaction with the target. Energy reaching the Earth's surface is reflected, absorbed, or transmitted depending on the target's properties. The reflected energy carries the spectral signature of the object. In active systems, the transmitted pulse is backscattered by the target, and the return time and intensity give information on distance and surface roughness.

4. Sensing by the sensor. A remote sensor (e.g., a multispectral scanner, radiometer, or radar) collects the reflected or emitted radiation, separates it into spectral bands using filters or dispersing elements, and converts it into an electrical signal. The signal is recorded as a digital number (DN) proportional to radiance.

5. Transmission, processing, and interpretation. The recorded data are transmitted to ground stations, corrected for atmospheric and geometric distortions, and analysed. The calibrated radiance is related to surface properties through the reflectance or emissivity of the target, allowing identification and mapping of features.

In summary, remote sensing operates on the principle that different objects interact differently with EM radiation. Energy from a natural or artificial source travels through the atmosphere, interacts with the target, and the reflected or emitted radiation is detected by a sensor, recorded, and interpreted to extract information about the Earth's surface without direct contact.

Physical and Functional Components of GIS

A Geographic Information System (GIS) is best understood through its physical components (the tangible parts that must be assembled) and its functional components (the operations or capabilities the system performs).

1. Physical Components of GIS

These are the five basic building blocks required to make a GIS work:

(a) Hardware The computer platform on which GIS software runs. It includes:

• Input devices — digitizers, scanners, GPS receivers, keyboards.
• Processing unit — CPU, RAM, and storage of the workstation or server.
• Output devices — plotters, printers, monitors, and display screens.

(b) Software The programs that provide the tools to store, analyse, and display geographic data. It includes the GIS package itself (e.g., ArcGIS, QGIS), the operating system, and supporting database and communication software.

(c) Data The most expensive and central component. Geographic data consists of spatial data (location, shape, topology) and attribute data (descriptive characteristics). Data may be in vector or raster form and must be accurate, current, and complete.

(d) People The users who operate the system — technicians, analysts, decision-makers, and end users. People are the most important component because they design, manage, and interpret the system.

(e) Methods / Procedures The well-defined rules and workflows by which data is entered, validated, analysed, and presented. These ensure consistency and reliability of results.

2. Functional Components of GIS

These describe what the system does:

(a) Data Input and Capture Conversion of analogue maps, field survey data, GPS points, and satellite images into digital form through digitizing, scanning, or direct import.

(b) Data Storage and Management Organisation of spatial and attribute data in files or a DBMS (spatial database), with proper linking between geometry and attributes.

(c) Data Manipulation and Analysis Includes:

• Overlay (union, intersect, identity)
• Buffer generation
• Network analysis (routing, allocation)
• Terrain analysis (slope, aspect, viewshed)
• Spatial query and statistics

(d) Data Output and Visualisation Production of maps, charts, reports, and 3D visualisations for presentation and decision-making.

(e) Data Updating and Management Continuous revision, editing, and quality control of the database over time.

---

Method of Defining Projection and Coordinate System in GIS

Every spatial dataset must be georeferenced to a common reference frame so that layers align correctly. This is done in two stages:

Step 1 — Define the Geographic Coordinate System (GCS)

The GCS fixes positions on the curved earth using latitude and longitude. It is defined by:

• Datum — the mathematical model of the earth's shape and its position relative to the earth's centre. Examples: WGS 84, Everest 1830 (Nepal), NAD 83.
• Prime meridian and angular unit (usually degrees).

In GIS this is set through the Data Frame Properties → Coordinate System tab, or by using the Define Projection tool when the data has no or wrong spatial reference.

Step 2: Define the Projected Coordinate System (PCS)

Because a curved surface cannot be flattened without distortion, a map projection is applied to convert latitude/longitude into planar X, Y coordinates. A PCS is defined by:

• The GCS/datum on which it is based.
• The projection type (e.g., Transverse Mercator, Lambert Conformal Conic, UTM).
• Projection parameters — central meridian, latitude of origin, standard parallels, scale factor, false easting and false northing.

For example, Nepal commonly uses the Modified Universal Transverse Mercator (MUTM) based on the Everest 1830 datum, with a central meridian of 84°E and a scale factor of 0.9999.

Step 3 — Assigning the Coordinate System in GIS

1. Define Projection — used when the data has unknown or incorrect projection. It only writes the metadata; it does not change coordinates.
2. Project (Project Raster / Project Feature) — used to transform data from one coordinate system to another, recalculating coordinates mathematically.
3. On-the-fly projection — GIS software automatically reprojects layers to the data frame's coordinate system for display, without altering the source data.

Step 4 — Verification

After defining or projecting, the layer is checked against a known reference layer (e.g., a base map or control points) to confirm correct alignment. Any mismatch indicates a wrong datum or parameter and must be corrected before analysis.