Statistics for Decision Makers - 01.00 - Outline

From Training Material
Revision as of 12:39, 12 June 2014 by Bernard Szlachta (talk | contribs) (→‎Statistics…。)
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to navigation Jump to search
title
Statistics for Decision Makers
author
Bernard Szlachta (NobleProg Ltd) bs@nobleprog.co.uk

Statistics…。

"There are three kinds of lies
lies, damned lies, and statistics."
  • Is it just yet another statistic?
  • Or is it just a lie?
  • It is just a lie!

What statistics can offer to Decision Makers。

  • Descriptive Statistics
    • Basic statistics - which of the statistics (e.g. median, average, percentiles etc...) are more relevant to different distributions
    • Graphs - significance of getting it right (e.g. how the way the graph is created reflects the decision)
    • Variable types - what variables are easier to deal with
    • Ceteris paribus, things are always in motion
    • Third variable problem - how to find the real influencer
  • Inferential Statistics
    • Probability value - what is the meaning of the P-value
    • Repeated experiment - how to interpret repeated experiment results
    • Data collection - you can minimize bias, but not get rid of it
    • Understanding confidence levels

Statistical Thinking。

  • Decision making with limited information
    • How to check how much information is enough
    • Prioritizing goals based on probability and potential return (benefit/cost ratio, decision trees)
  • How errors add up
    • The Butterfly Effect
  • The Cassandra Problem - how to measure a forecast if the course of action has changed
    • How decisions make forecasting outdated
  • Forecasting - methods and practicality
    • ARIMA
    • Why naive forecasts are usually more responsive
    • How far a forecast should look into the past
    • Why more data can mean worse forecasts

Statistical Methods that are useful for Decision Makers。

* Describing Bivariate Data
    • Univariate data and bivariate data
  • Probability
    • Why things differ each time we measure them
  • Normal Distributions and normally distributed errors
  • Estimation
    • Independent sources of information and degrees of freedom
  • The logic of Hypothesis Testing
    • What can be proven, and why it is always the opposite of what we want (Falsification)
    • Interpreting the results of Hypothesis Testing
    • Testing Means
  • Power
    • How to determine a good (and cheap) sample size
    • False positives and false negatives and why it is always a trade-off

Statistics。

Descriptive (without extrapolating)
  • Used for:
    • Comparing populations
    • Tracking trends (e.g. profit)
Inferential (uses probability to extrapolate)
  • Using samples to infer what a population looks like
  • Using past data to infer the future (forecasting)

Probability。

  • Frequentist approach
  • Probabilities can be combined (addition, multiplication)
  • Usually probabilities are conditional (probability of A assuming B)

Statistical Errors。

  • Error of Measurement
    • What is the margin of error of the device we use to measure?
  • Bias of Sampling Method
    • Is the sample really representative?
  • False Detection (false positive, Type I)
    • What is the probability of finding a difference where in reality there is no difference?
  • Missed Detection (false negative, Type II)
    • What is the probability of not finding the difference where there is one?
  • Forecast Error
    • How much the forecast differs from the real thing

Control Charts。

  • Early warning system that something has changed
  • Is the average page response time slowing down or were the last samples just unusually slow?
  • Is the response rate from customers exceptionally low?
  • Should we stop and investigate the increased response time or ignore it?


Auto-regression。

  • Are changes in the past related to changes in the future?
  • Are sales in March always lower on average than in other months of the year?