- make causality a separate lecture 10, add on diff-in-diff with cannabis example.
- add example coverage calculations Lect 2 or ARIMA, eventually move part of the material from l02 to a separate lecture on Model evaluation and forecasting. forecast::accuracy(). caret::postResample(obs = y_test, pred = y_dnn) Overfitting. B&D Ch. 9 Remember that all predictive inference is based on the assumptions that the relationships between the variables and their dynamics will be the same in the future.
- add a lecture on panel data analysis
- expand on wavelets, add example of acoustic data
- appendix or lecture on changepoint analysis
- GLM or another generalized model with classification example for fisheries or ecology.
- local stationarity, warping, time motifs, graph representation of time series
- from NOAA project with G.N.: "Environmental Statistics 2: identification of different types of rare events in time series (appendix?), time-series cross-validation (l 12?), and GAMLSS (lect 12 on TSREG2 - done; lect 3 - done)"
a.k.a. changepoint dataset heteroskedasticity homoskedasticity hyperparameter nondeterministic nonlinear nonnegative nonparametric nonstationarity non-existent non-monotonic non-normal non-overlapping non-seasonal scatterplot vs.
\boldsymbol \dots (not \ldots or \cdots)
Use italics for highlights in text, not bold.
Use 'single quotes' in text whenever possible.
#| code-fold: false
Space and capital letter after a comment sign:
Cite @Brockwell:Davis:2002 or [@Brockwell:Davis:2002] or [@Rebane:Pearl:1987;@Pearl:2009]
Recall the classical decomposition $$ Y_t = M_t + S_t + \epsilon_t, $${#eq-trseas}
model as @eq-trseas is
fig-height use default (5) for 1-2 plots per row #| fig-height: 3 for 3 plots per row #| fig-height: 7 for decompose or 2 rows #| fig-height: 9 for 3 rows
#| label: fig-shampoo
#| fig-cap: "Monthly shampoo sales over three years and a corresponding sample ACF."
p1 <- forecast::autoplot(shampoo) +
xlab("Year") +
ylab("Sales") +
theme_light()
p2 <- forecast::ggAcf(shampoo) +
ggtitle("") +
xlab("Lag (months)") +
theme_light()
p1 + p2 +
plot_annotation(tag_levels = 'A') &
theme_light()
::: {.callout-note} text :::
::: {.callout-note icon=false}
text :::
from 0 to |
from |
from |
from |
from |
---|---|---|---|---|
Reject |
Neither accept |
Do not reject |
Neither accept |
Reject |
: Regions of rejection of the null hypothesis for the Durbin--Watson test {#tbl-DW}
Collapse All — Alt+O. Expand All — Shift+Alt+O.