Summary

• Hypothesis testing is a key to quantify the differences in data.

• Null hypothesis are usually ‘no effects’ or ‘no differences’.

• T-test, ANOVA, and non-parametric tests could be used in various conditions.

• T-test and ANOVA were designed with some assumptions and base distribution (t-distribution and F-distribution), which are prefered if the assumptions are met.

• If the assumptions are violated, the statistical estimation is compromised and the accuracy is unknown.

• Non-parametric tests can be used in that situation.

How to think statistically?¶

Some key points:

• How to measure data quantitatively (or qualitatively)

  • nominal, ordinal, interval, ratio

  • binary, numerical, categorical

• Are them different? Are them same?

  • designing the hypothesis

  • how to formulate a question

• Are them related?

  • correlation, regression (not covered)

• How to test/assess the hypothesis?

  • parametric, non-parametric

  • the assumptions behind each method

• How to read the results (interpretation)?

  • significant at what level?

  • effect size?

  • how confidence are we?

• Relating to the big picture

  • what this means in your big-aim / Reseearch Questions (RQ)

  • how visualisation can help linking back to your study

How to use these techniques and thinking in spatial data?¶

Strategies to investigate spatial data differences¶

• individual level comparison

  • directly analyzing the data table

  • transactions between flat types, between regions, between time, ...

  • each data point is considered independent from other data point

  • use statistical methods

• spatially aggregated comparison

  • aggregate the data by spatial units (subzone, planning area)

  • the average price for 4 rooms for every subzone, ...

  • each spatial unit is considered independent from other spatial unit

  • use statistical methods

• consider spatial relationship (nearby) effects

  • spatial relationship between data points or spatial units

  • the prices of the same type of house unit within the neighborhood (a search distance)...

  • data points or spatial units are consider spatially dependent

  • use spatial statistic methods