Quantitative analysis Flashcards
What are the main tasks of multiplre regression and what is it used to do?
What are the five assumptions for MLR?
What is a scatterplot matrix and when you do/dont want to see linear relationships? What does it help do?
Helps identifies outliers and visualise homoskedasicty
What does non-constant and autocorrelated variance look like? Standardized vs normal distribtion?
What is the coefficient of determination, what happens to R^2 as IVs are added, what does R^2 not tell us? Adjusted R^2 formula?
Adjusted R^2 what it does and does not tell us? What is AIC and BIC? When should each be used?
AIC better for prediction, BIC better for goodness of fit. A decrease in adjusted R^2 from adding an IV means that the IV has no explanatory value
How to test join coefficients? formula for F-stat?
What is the difference between standard error and standard deviation?
What are the principles for good regression model specification?
What are the four failures in regression functional form and what does homoskedastic and heteroskedastic mean?
What are the consequences from failiure in regression functional form: unconditional heteroscedasticity and conditional heteroskedasticity, how to test for CH?
How to correct for CH, what is serial correlation and positive and negative sc?
How does SC affect tests and how to test for serial correlation?
How to correct SC, what is multicollinearity and how to detect it?
What are the solutions of MC? Explain the assumption, violation, detection and solution to Homoskedastic errors, Independence of observations and independance of independent variables
What are the types of influential data points, how to detect/leverage them,
Method for detecting outliers - STudentized?
What is the D-measure of influence?
Determining the results of DW test for serial correlation? How to calculate DW statistic and what does a result of 0 imply?
If test statistic is lower the Dl then reject null of no positive serial correlation, if between Dl and Du its inconclusive and if greater than Du then fail to rejet the null of no positive serial correlation. DW = 2(1-r) where r = correlation between residuals, therfore perfectly positive correlated residuals have a DW of 0
What is the null and alternative Hypothesis for f-test of joint significance?
Null, H0 = b1=b2=b3…=0,alternative is that one of them is not and rejecting the null means the variables are jointly signifncant but not necessarily all indiviually significant.
Summary for measure of influence, whether it can be used for independant or dependent, the process and the method for the basis for being influential
Describe a dummy variable, its uses and an intercept and slope dummy?
Adding dummy variables and interpeting them
What are qualitative variables, describe the logit model
What are the properties of logistic regression?
What is the relationship between Log-odds with changes in IVs and what is the goodness of fit test for them
What is the log-liklihood criteria for goodness of fit?
Multiply the LL null by the log ratio -2(llnull(restricted) 1-llnull2(unrestricted) and whichever is closest to zero (larger) is the better fit
What is the difference between time series and a causal relationship?
What is a linear trend model?
Describe the log-linear trend model?
What is an auto-regressive time series model and what must the mean, variance and covariance between observations be?
How to deal with serial correlation in AR models, what is the test statistic, what are the three steps?
Explain mean reversion, when the dependant variable will increase, decrease, stay the same and what is the chain rule of forecasting?
How to explain model forecast error, explain in-sample vs out of sample forecast errors, whats is RMSE?
Regression coefficients may vary based on what two factors?
What is a random walk, what is a unit root, what do RW not have?
What is first differencing, what does it achieve, what we can conclude from doing it?
What is a RW with a drift, what is the unit root test of non-stationarity, describe how the DF test is done?
What are Seasonal models and ARCH models?
What happens if any TS in LR has a unit root , when testing each TS for unit roots if both reject null? What if reject for IV but not for DV? What if reverse? What if both have unit root? what if both have unit root but not cointegrated?
If combinations of TS with unit roots offer valid solutions, and how to test for cointegration?
What can be done if more than one IV has a unit root?
What is supervised learning?
What is unsupervised learning?
What is a deep learning and summary what each type of learning goes with what problem?
What is overfitting, how can data sets be partitioned, what leads to poor prediction?
What does high bias err, high variance error and good trade off between the two look like? Which type of model are more susceptible to which error?
Two methods to prevent overfitting? What is k-fold test validation?
What is penalized regression?
What is a support vector machine?
What are classification and regression trees?
What is ensemble learning and random forrests?
Describe principal component analysis?
Describe k-means and hiearaccheal clustering.
What is a dendogram?
What is a neural network?
What are deep learning nets and reinforcement learning?
What are the 4 Vs of how big data differs from traditional data?
What are the steps in data analysis projects for traditional and textual data?
What are the common errors associated with normal (structure) data?
What are the procedures included in preparing or pre-processing structured data?
What are the two types of scaling done to structured data? In transforming unstructured to structured data what needs to be removed?
Describe the process of cleansing unstructured word data and normalising text data?
Describe data exploration for structured data?
Describe data exploration for unstructured data? What are the methods of eliminating the noisy features?
- Mutual information how much information a token contributes to a class.
Describe feature engineering?
What is the first step in Model training?
Describe the performance evaluation step and what is precision and recall?
What is the ROC, TPR, FPR, RMSE and describe tuning?
What is standard error and t-stat for autocorrelations?
When are two series cointegrated?
How to calculate periodic pension expense
How to calculate retirement benefits paid
What is recorded in OCI under GAAP and how to calculate
