everything but art

ohayou! bonjour! guten morgen!

started the day earlier than usual - 09.09!

managed to finish my language dailies in less than half an hour.

french: elementary a2 - 1.7 developing fluency

french	english
J'ai fait le ménage et après, j'ai regardée la télé.	I did some housework and then I watched TV.
Je suis restée à la maison pendant le week-end.	I stayed home over the weekend.
Je suis sortie avec des amis toute la soirée.	I went out with friends all evening.

german: elementary a2 - 1.5 developing fluency

german	english
Das Wetter war schön. Ich war im Park.	The weather was nice. I was in the park.
Ich hatte Besuch. Meine Schwester war hier.	I had a visitor. My sister was here.
Ich war am Wochenende im Park.	I was in the park at the weekend.
Ich war kaputt. Ich war zu Hause.	I was knackered. I was at home.

regression? like, in orv?

i finished the datacamp course 'introduction to regression with statsmodels in python' just in time for lunch (12.27). here are my notes for the course:

chapter 1: simple linear regression modeling

• regression
  • statistical models that help explore relationship between a response variable and explanatory variables
  • if given the values of explanatory variables, values of response variables can be predicted
• response variable
  • y variable
  • dependent variable
  • the variable you want to predict
• explanatory variable
  • x variable
  • independent variable
  • variables that explain how the response variable will change
• linear regression
  • when the response variable is numeric
• logistic regression
  • when the response variable is logical (either True or False)
• simple linear / logistics regression
  • only one explanatory variable
• a scatterplot can be used to visualise pairs of variables
• python packages for regression
  • statsmodels - for insight
  • scikit-learn - for prediction
• a histogram can be used to visualise the relationship between numerical and categorical variables

chapter 2: predictions and model objects

• the predicting question: if i set the explanatory variables to these values, what value would the response variable have?
• extrapolating
  • making predictions outside the range of observed data
• fitted values (.fittedvalues attribute) - predictions on original dataset
• residuals (.resid attribute) - actual response values minus predicted response values (how much the model missed by)
• response value = fitted value + residual
• regression to the mean
  • when residuals exist due to problems in model and fundamental randomness
  • extreme cases are often due to randomness
  • extreme cases don’t persist over time - will eventually look like average cases
• to fit a linear regression model, may need to transform the explanatory or response variable if they do not give a straight line

chapter 3: assessing model fit

• coefficient of determination (r-squared or R-squared)
  • the proportion of the variance in the response variable that is predictable from the explanatory variable
  • 1 is a perfect fit
  • 0 means worst possible fit
  • correlation squared - for simple linear regression
• residual standard error
  • roughly a measure of the typical size of the residuals - how much the predictions are typically wrong
• leverage - a measure of how extreme the explanatory variable values are
• influence - measures how much the model would change if you left observation out of the dataset when modeling

chapter 4: simple logistics regression modeling

• logistic regression
  • type of generalised linear model
  • user when the response variable is logical
  • follows s-shaped curve
• odds ratio - the probability of something happening divided by the probability that is doesn’t
  • odds_ratio = probability / (1-probability)
• four outcomes to a logical response variable
  • predicted false, actual false - correct
  • predicted false, actual true - false negative
  • predicted true, actual false - false positive
  • predicted true, actual true - correct
  • a confusion matrix - the counts of each outcome
• accuracy of model
  • the proportion of correct predictions
  • accuracy = tn+tp / tn+fn+fp+tp
• sensitivity of model
  • the proportion of true positives
  • accuracy = tp / fn+tp
• specificity of model
  • the proportion of true negatives
  • tn / tn+fp

blog checkpoints

today's catchup for blog posts was for the 29th, 30th, and 31st may, as well as yesterday's blog, the 4th june. here are the times i was able to finish them, starting from about 16.30:

• 17.19 finished 29th may blog page
• 18.11 finished 30th may blog page
• 18.32 finished 31st may blog page
• 19.09 finished yesterday's blog page!!