6,803 reputation
11142
bio website sap.com/belgique/solution/…
location Switzerland
age 39
visits member for 4 years, 3 months
seen 4 hours ago

During the day, I create software to forecast retail sales and calculate order proposals. Lots of time series, with an emphasis on fast, automatic and robust data cleansing and forecasting. I'm an elected director for the International Institute of Forecasters and an Associate Editor for their practitioner-oriented forecasting journal Foresight. At night, I switch hats and do inferential statistics for academic clinical and biological psychology. My main tool is the statistical computing environment R.

profile for Stephan Kolassa on Stack Exchange, a network of free, community-driven Q&A sites


2d
comment Hourly predictions using time series
@NickCox: fully agree. Dummy coding months (or hours) is really the perennial Worst Practice of discretizing continuous predictors.
2d
revised Hourly predictions using time series
formatting and tags
2d
answered Hourly predictions using time series
2d
comment Is high AIC a bad feature of the model?
+1. You can't interpret the absolute value of AIC. (Different software packages may well give completely different AICs on the same data for the same model.) What you can interpret is the difference in AIC between different models applied to the same data. Burnham & Anderson give some rough rules of thumb: a difference of 2 means that both models are essentially equally good, 5 means that the model with the lower AIC is a bit better, 10 is pretty strong evidence that the lower AIC model is better.
Dec
18
comment How can we compute cumulative change rates for time series data?
What do you mean by "cumulative rate change"? Can you edit your question to give an example?
Dec
18
comment How to predict/find a value based on historical data using statistics?
In case this is time series related, I like to recommend this free open source online forecasting textbook. Also, voting to close as "unclear what you are asking". Please consider editing you question to clarify it.
Dec
17
revised Good practice for statistical analysis in a business environment
added consulting tag
Dec
17
comment Good practice for statistical analysis in a business environment
You may want to look into knitr.
Dec
17
comment What is a good mouse cursor to point at objects in diagrams?
This question appears to be off-topic because it is about UI/UX. It should be migrated to ux.stackexchange.com
Dec
17
revised Selecting the best (or more suitable to the user/client) output from a set of forecasts
removed R tag
Dec
17
comment Selecting the best (or more suitable to the user/client) output from a set of forecasts
Suppose actual sales are 2,2,2,2,2 (since we are assuming that the actual best forecast is non-seasonal). The non-seasonal forecast is 2,2,2,2,2. The seasonal forecast is 2,3,2,1,2. Both forecasts have the correct average of 2. If you plan shelf space (or inventory) based on the seasonal forecast, you have excess space in week 2 and too little space in week 4. If you don't like the flat actual, you can simulate random actuals with no seasonality around some constant mean and check the implications of seasonal and non-seasonal forecasts on a series that you know is non-seasonal.
Dec
17
comment Selecting the best (or more suitable to the user/client) output from a set of forecasts
Thanks. I still don't understand. Why does a flat line not allow you to estimate shelf space requirements? (They will simply not change over time.) Why is a seasonal forecast better for this although it is a worse forecast? It may be worthwhile to think some more about why exactly you want seasonality. (Incidentally, I have been forecasting retail sales for almost a decade now. Often, clients don't like a flat line because it doesn't look "sophisticated" enough - "why did I pay a lot of money to get a flat line?" - although it is the best forecast and gives the best inventory position.)
Dec
17
comment Selecting the best (or more suitable to the user/client) output from a set of forecasts
Could you edit your question to give an example where a seasonal forecast is better for making decisions than a better nonseasonal forecast? I have to admit that I don't understand that.
Dec
17
answered Selecting the best (or more suitable to the user/client) output from a set of forecasts
Dec
17
comment How to measure consistency or simiarity between two data sets?
Cosine similarity is often used to compare vectors.
Dec
17
comment Calculating spline curve with custom knot positions
I'm afraid this really exceeds what we can use SE comments for... Do you know matrix algebra? (If not, you'll not get far.) The fit plotted above is really just a matrix (a column of ones, then d$splines), multiplied by a vector for coefficients (in model). Best to learn about OLS (Wikipedia), then understand how ns creates the design matrix of splines, for which the Harrell book (section 2.4.4) is very useful (though very terse). The splines are related to powers of x, but they are not the same, so you don't have relationships between y and x^n. Good luck!
Dec
17
comment Test for Normality
In addition to @Glen_b's excellent comment and Aksakal's equally excellent answer, note that even for continuous distributions, K-S requires that the mean and sd be known beforehand, not estimated from the data. This essentially makes the K-S test useless. "The Kolmogorov-Smirnov test is only a historical curiosity. It should never be used." (D'Agostino in d'Agostino & Stephens, eds., 1986). If at all, use Shapiro-Wilks instead.
Dec
16
comment Calculating spline curve with custom knot positions
model is a fitted Ordinary Least Squares model (lm stands for "linear model"), based on d$splines (and an implicit leading intercept column) as the design matrix. If you simply type model, you get the estimated regression coefficients.
Dec
16
comment Calculating spline curve with custom knot positions
ns calculates piecewise cubic splines, not quadratic ones. In addition, they are natural in the sense that they are linear outside the knots. Without having gone through the details, I think this means that the answer you linked to is not applicable here, i.e., you can't reconstruct d$splines using that answer. The reference in ?ns may be helpful, as may be the splines section in Harrell's Regression Modeling Strategies.
Dec
16
answered Calculating spline curve with custom knot positions