Statistics in Scientific Research

Study Notes

Statistics are crucial for summarizing and graphically representing data in scientific research.
Data quality is essential for informed decision-making, accurate models, and effective policy formulation.
Key dimensions of data quality include accuracy, consistency, completeness, timeliness, and relevance.
Data is raw facts, numbers, or symbols, while information is processed data with context and meaning.
Statistical methods are vital for analyzing environmental data, predicting outcomes, and developing environmental policies.
System analysis helps understand complex environmental issues by modelling interconnected factors.
Descriptive statistics summarize data (mean, median, mode, standard deviation).
Inferential statistics draw conclusions about populations based on sample data (hypothesis testing, regression).
Statistical models like regression analysis predict trends (e.g., temperature rise), time series analysis studies changes over time (e.g., annual CO2 emissions), and multivariate analysis analyzes multiple factors impacting ecosystems.
Biostatistics is the science that develops and applies appropriate methods for collecting, presenting, analyzing and interpreting data for making decisions. Data sources include sensors (air, water, soil), satellite data, and citizen science.
Different statistical procedures allow for diverse research questions (describing, correlational, or experimental).

Descriptive questions focus on describing the characteristics of a variable (What does X look like?).
Correlational questions examine the relationship between variables (Is there an association between X and Y?).
Experimental questions investigate if changes to one variable cause changes in another (Do changes in X cause changes in Y?).

Data types include categorical (e.g., male/female, blood type) and numerical (e.g., weight, number of cells).
Categorical data can be nominal (name/label) or ordinal (rank order), while numerical data can be interval (equal intervals) or ratio (equal intervals and absolute zero).
Nominal data has no inherent ranking (e.g., blood type),
Ordinal data has a ranking, but the differences between values aren't necessarily equal (e.g., survey ratings).
Interval data has equal intervals between values, but there may be no true zero (e.g., temperature).
Ratio data has meaningful zero and equal intervals between values (e.g., weight).

Variables can be quantitative (numerical) or qualitative (categorical).
Quantitative variables can be continuous (e.g., height) or discrete (e.g., number of students).
Qualitative variables can be nominal (e.g., colors), or ordinal (e.g. rankings).

Data can be presented numerically (tables), graphically (charts), or mathematically (summary statistics).
Examples of graphical presentations include line graphs, frequency polygons, histograms, bar charts, scatter plots, pie charts, and statistical maps.
Example of mathematical presentations include measures of location (e.g., mean, median, mode) and measures of variability (e.g., range, standard deviation).

Statistical procedures include descriptive statistics (organizing and summarizing data) and inferential statistics (drawing inferences from data).
Descriptive statistics use methods like frequencies, contingency tables, measures of central tendency, and measures of variability. Graphical summaries include bar graphs, histograms, scatterplots, and time series plots.
Important considerations include data quality, clear labeling of axes, and appropriate choice of graph or chart type.

Label axes and provide a heading for clarity.
Vertical axes should start at zero for comparing relative sizes.
Remove unnecessary clutter.
Consider resizing axes to remove extra white space.
Use careful bars for relative percentages to avoid misrepresentation.
If displaying multiple groups on a single graph, consider using relative frequencies or separate graphs.

### Other Guidelines

Pie Charts: For categorical data with proportions.
Doughnut Charts: Similar to pie charts, good for comparing subsets.
Scatterplots: for showing a relationship between two numerical variables.
Time series plots: Numerical data plotted over time.
Bar Charts/ Histograms: Representing frequencies of categories or intervals, respectively.

Normal distributions are approximately symmetric, bell-shaped.
Skewed distributions are asymmetrical (positive skew or negative skew).
Bimodal distributions have two peaks, while uniform distributions have flat peaks.
Outliers are extreme data values.