Forecasting
Gio sells gelato and collected sales data for the past few days. Gio wants to use the Naive method to forecast. Calculate the Naive forecast for Day 6. Day 1 to 5 sales= 90,97,92,95,92
92
Which of the following are Quantitative Methods 1. Naive Methods 2. Exective Opinion - true 3. Delphi Method - true 4. Market Research - true 5. Sales Force Opinion - true
Naive method is NOT a qualitative method. Naive method utilizes past quantitative data. See the next video on the use of Naive method.
1. The model predicts we would reach $1 million in sales, and we should believe it. Woo-Hoo! 2. We should use the previous linear trend, anyways, because it's R-squared is lower and thus better. 3. This is such a huge number our computer will blow up. 4. None of these things above are correct.
(1) is incorrect because we would be over-extrapolating. (2) is incorrect because models with higher R-squared are better. And once again, nothing we do in this class will explode anything, so (3) is incorrect. (4) is the only correct statement.
Which of the following statements is True? 1. If a linear trend has a correlation of 0.75 and an associative forecast for the same sales data has a correlation of -.80, the linear trend has better predictive ability because .75 > -.80 2. If a linear trend has an R-squared value of 70% and an associative forecast for the same sales data has an R-squared of 60%, then the linear trend has a better predictive ability because 70% > 60%. 3. Let's say a friend tells you he has an associative forecast with an R-squared value of 110%, His modle must be the best, right? 4. An associative forecast predicting a person's salary given their age has a high R-squared value meaning that younger people should sue employers for discrimination. 5. You are going to have to calculate a and b by hand in this class.
1) FALSE. Correlation has both strength and direction. The associative forecast with a correlation of -.80 is stronger than the linear trend's of .75. Thing in terms of absolute values (or just square the correlations to work with R-squared values). 2) True. The higher R-squared, the better predictive ability. 3) FALSE. Your friend is mistaken- there's no way R-squared can be greater than 100% 4) FALSE. The model only shows an association between salary and age, and correlation is not a guarantee of causation. There could be a lurking variable: many of the older people may have been in the workforce longer and may accumulated more promotions and pay raises. To truly see if there is age discrimination, we'd need to compare salaries for people with the same amount of job experience and education but different ages. (For some fields such as tech, there's evidence that reverse age discrimination may exist.) 5) FALSE. We will plug the data into Excel to get a and b.
Continuing with the lemonade data, lets see if we can calculate forecast accuracy for the Naive and all of the other smoothing methods defined in previous PRACTICEs.1) Which forecasting methods can we not calculate MAD for? Why?2) For the ones that we can calculate MAD for, which seems to be the most accurate?3) Is that method always going to be the best one for all stable time series?
1) Since we only have 4 days of sales data, we cannot calculate MAD for either the 4-period MA or the 4-period WMA. 2) See the image below for the calcuations for the remaining 3 techniques. Based on the data we have, the 3-period MA outperforms both the Naive and the Exponential Smoothing, because its MAD is lower. 3) However, that doesn't mean the 3-period MA is always the most accurate technique for all stable time series.
Which of these statements about time-series methods are false and why? 1. Any quantitative data are time-series data. 2. Percentage of vaccinated adults in urban areas are time-series data 3. Weights and heights of athletes are time-series data. 4. You had counted the number of patients arriving at an emergency room between a) 8 am until 10 am b) 10 am and 10:30 am c) 10:30 am and 12 pm d) 12 pm and 2 pm. These are time-series data.
1- false. Not all quantitative data are time-series data. 2. false. Although these are quantitative data, they are NOT time-series data because there is no time component. 3-false. Although these are quantitative data, they are NOT time-series data because there is no time component. 4- false. Although there is time component, these data are NOT time-series data because the data were not collected at equal-time intervals. The first time interval is 2 hours long whereas the second and third time intervals are 30 minutes and 90 minutes. The time intervals have to be all in equal length.
Which of these statements about naive forecasts are false and why? 1. The simple naive forecast requires only one data point. 2. Naive forecasts are simple to compute: we write the formula as Ft= A t 3. Because naive forecasts don't tend to be as accurate as other forecasting techniques, they have no use whatsoever. 4. If you do a naive forecast for your boss, you will have to spend a lot of time explaining what you did, as they are hard to understand.
1- true. 2. false. While they are simple, the formula is Ft= A t-1 Rembember, without clairvoyance we can only forecast given prior data. 3-false. While it is true that naive forecasts don't tend to be highly accurate, we can use their relative inaccuracy as a benchmark to measure other forecasts against. If they don't do as well as the naive, then they may not be appropriate for the data in question. 4- false (I hope). Naive forecasts are easy to understand. If your boss can't... maybe it's time to look for a new job?
Mai sells makeup online and has noticed that demand for her products, glittery lip gloss, has varied wildly over the past few months she has been selling it. She's been doing some online advertising and thinks there might be a relationship between how much she's spending each week on ads and her weekly lip gloss sales, so she's used Excel to create an Associative Forecast: where the intercept is -20 and the slope is 2.4 per $ spent advertising. Use this information to forecast lip gloss sales in week 10, when she will be spending $100 on advertising.
2.4*100-20 =220
What is the standard deviation for the following sample of numbers? 8.5, 1.8, 8.3, 6.0, 6.4 Round your answer to 4 decimals.
2.6991
Mai sells makeup online and has noticed that demand for her products, glittery lip gloss, has been growing somewhat linearly since she introduced it 9 weeks ago. She's done a linear trend analysis to forecast weekly sales in Excel, where the intercept is 20 and the slope is 14.1/week. Use this information to forecast demand for week 10.
20+14.1*10= 161
Jeff Brown has collected sales data for his cupcake stand. Calculate the moving average forecast for period 13 with n = 4. Round your answers to 2 decimals. Time = 1 to 12 Sales = 48,41,37,32,36,34,43,52,60,48,41.9,29.7
44.9
Jeff Brown has collected sales data for his cupcake stand. Calculate the moving average forecast with n = 4. Time = 1 to 12 Sales = 48,41,37,32,36,31,43,52,62,47,44,30
45.75
According to the syllabus for a section of DS 412, the weights for the quizzes, the midterm, and the final are, respectively, 10%, 50%, and 40%. If a student has an average of 96 on the quizzes, 77 on the midterm and 69.3 on the final, how many percentage points are earned by that student at the end of the semester? Keep 2 decimal places.
75.82
Chico sells churros and has collected sales data for the past few days. Chico will use a two period moving average to forecast, but is concerned about forecast accuracy. Calculate MAD to 1 decimal place Sales = 100,100,99
99.5 - wrong 99.7 - wrong the right answer is 1
1. Excel will convert dates in September 2020 to numbers like 44,075 2. The intercept will have a really odd value, compared to if we reset our start date to 0 or 1. 3. he slope and R-squared will not be any different than if we had reset our start date to 0 or 1 4. Our computer will blow up.
Everything but (4) will happen. Nothing we do in this class will explode anything.
True or false: Naïve forecasting methods are not appropriate for use in business. Select one: True False
False
Qualitative methods are inferior to quantitative methods.
False. While it's true that qualitative methods are subjective, this does not necessarily make them inferior to quantitative methods. Qualitative methods have the ability to capture an ongoing change in the environment that quantitative methods may not be able to. Think of toilet paper sales during the pandemic. Past year's data (March 2019 - February 2020) was not helpful in predicting sales of toilet paper in March 2020
Say you have 4 days of daily sales data for your kid's lemonade stand: Monday: $20, Tuesday $25, Wednesday $21 & Thursday $24. Compute both the 3-period and 4-period moving average (MA) for Friday and all days earlier as appropriate. Also, does the data appear to be a stable time series?
For the 3 period MA, we'd start on Thursday( =($20+$25+$21)/3 = $22) and calculate Friday's forecast as = ($25+$21+$24)/3 = $23.33 For the 3 period MA, we'd start on Thursday( =($20+$25+$21)/3 = $22) and calculate Friday's forecast as = ($25+$21+$24)/3 = $23.33 For the 4-period MA, we can only calculate Friday's forecast: = ($20+$25+$21+$24)/4 = $22.5 Although 4 days isn't a lot of data to judge by, this seems to be a stable time series- no trend upwards (or down), and while every other day seems to result in a bit more sales than the others, it's probably too early to claim any seasonality effect. An example of an outlier would be if one day we sold more than $30- perhaps because of a heat wave.
Francois considers that the introduction of Frose' coincides with a recent heatwave. He's created a new model that uses maximum temperatures to predict sales, and Excel gives a slope of 1.5 (per degree Farenheit) and an intercept of -40. What do we predict today's demand for Frose' will be, if the temperature is expected to get up to 90 degrees F? How should we interpret the negative intercept?
The Frose' model would be: Yt = -40+ 1.5*F (where F is degrees Farenheit) at if F = 90, the model predicts -40+1.5*90 = 95 sales of Frose' The negative intercept means that if the temperature were 0 degrees the model predicts negative sales, which doesn't make any business sense. But realistically this would be a misuse of the model, as it was built with sales data from a heatwave, so using it to predict sales in cold weather is a different sort of over-extrapolation!
Francois has started selling Frose' (pink wine slushies) at his beach cafe and has noticed sales have been growing at an approximately linear rate. He has calculated a slope of 5/day and an intercept of 30. What does the model predict demand for Frose' will be, given today will be the 10th day of sales? Should he use this same method to predict what sales could be 100 days from now?
The Frose' model would be: Yt = 30+ 5*t at t = 10, the model predicts 30+5*10 = 80 sales of Frose' at t = 10+ 100 = 110, the model would predict sales to grow to 30+5*110 = 580 ! But Francois should realize that this would be an over-extrapolation!
Which of the following are qualitative forecasting methods? Select one or more: a.Delphi Method b.Customer survey c.Associative forecasting d.Exponential smoothing
a.Delphi Method b.Customer survey d.Exponential smoothing
Weighted moving average forecasting is suited particularly well for what kind(s) of data sets? Select one: a.Stable/stationary series b.None of the other choices c.With linear trend d.Seasonal data
a.Stable/stationary series
Onisha manages a group of apartment complexes and is trying to create a budget for next year. Below are the monthly expenses for the last three years, in thousands of dollars. How many seasons are there? How many seasonal relatives (indices) do we need to forecast? a.The pattern repeats monthly, so we need 12.00 indices. b.None of the other options. c.The pattern repeats every year so we need 3 indices, one for each year. (incorrect) d.There are 36 observations, so we need 36 indices.
a.The pattern repeats monthly, so we need 12.00 indices.
Consider the following graph. Which forecasting method(s) would not be appropriate to use? (double -curve graph) Select one or more: a.Trend analysis b.Exponential smoothing c.Seasonal analysis (incorrect) d.Moving average
b.Exponential smoothing d.Moving average a.Trend analysis
When comparing the accuracy of the simple Naive forecast to a 3-period Moving Average by calculating MAD, which statement is correct? Select one: a.The forecasting technique that has the higher MAD is more accurate. b.The forecasting technique that has the lower MAD is more accurate. c.A Moving Average forecast is always better than a Naive forecast, so you never need to calculate and compare MAD. d.None of the other choices
b.The forecasting technique that has the lower MAD is more accurate.
Happy Baby, a maker of baby foods, has found a high correlation between the aggregate company sales (in $100,000) and the number of births nationally the preceding year (one year prior). Suppose that the sales and the birth figures during the past eight years are (8 years, sales in 100,000 and U.S births in million) a."Births" is the independent variable and "Year" is the dependent variable. b."Sales" is the independent variable and "Year" is the dependent variable. c."Births" is the independent variable and "Sales" is the dependent variable. d."Sales" is the independent variable and "Births" is the dependent variable. (wrong) e."Year" is the independent variable and "Sales" is the dependent variable
c."Births" is the independent variable and "Sales" is the dependent variable. Independe variable - a factor that purposely change or in control in order to see what affect it has dependent variable - a variable (often denoted by y) whose value depends on that of another.
Given forecast errors of 4, 8, and -3, what is the mean absolute deviation? Select one: a.9 b.5 c.None of the above d.3 e.15
c.None of the above - wrong d.3 - wrong a. 9- wrong the right answer is b. 5
Spears Rowbuck has recorded the following sales figures. What is the appropriate season length? (4 weeks in 5 days) a.There are a total of 20 observations, so we need to calculate 20 indices. b.None of the other options. c.There is a daily seasonal pattern, so calculate 5.0000 indices. d.There are 4 weeks, so we need to calculate an index for each week. Clear my choice
c.There is a daily seasonal pattern, so calculate 5.0000 indices.
Consider the following graph. Which forecasting method(s) would not be appropriate to use? (trending graph) a.a weighted moving average (incorrect) b.Moving average (incorrect) c.Trend analysis d.Exponential smoothing (incorrect)
c.Trend analysis
The seasonal indeces (relatives) for the months of January, February, and March are given as 1.10, 0.99, and 0.90, respectively. The sales of coffee cups in January, February, and March of 2012 were respectively 88, 99, and 108. Compute the deseasonalized sales data accordingly. Select one: a.88, 99, 108 b.Enough information is not given. c.96.8, 98.01, 97.2 d.80, 100, 120
d. 80, 100, 120
You are told to forecast demand for period 6 using a weighted moving average method with n = 3 and weights 0.4, 0.3, and 0.2 (from most recent to third most recent). Which of the following should you multiply with the weight 0.4 to get the forecast? Select one: a.A2 (incorrect) b.A6 (incorrect) c.A1 (incorrect) d.A5 e.A3 (incorrect) f.None of the other options. (incorrect) g.A4 (incorrect)
d.A5
Which of the following are qualitative forecasting methods? Select one or more: a.Customer survey b.Associative forecasting c.Exponential smoothing d.Delphi Method
d.Delphi Method a.Customer survey
Consider the following graph. Which forecasting method(s) would not be appropriate to use? (decrease trend graph) Select one or more: a.Trend analysis b.Exponential smoothing c.Moving average d.Seasonal analysis
d.Seasonal analysis b.Exponential smoothing c.Moving average
If you do a naive forecast for your boss, you will have to spend a lot of time explaining what you did, as they are hard to understand.
false (I hope). Naive forecasts easy to understand. If your boss can't... maybe it's time to look for a new job?
If a forecast isn't 100% accurate it's worthless!
false. No real forecast will be 100% accurate. Forecasts just need to be good enough to be useful. How good is "good enough?" We will discuss that later!
Because naive forecasts don't tend to be as accurate as other forecasting techniques, they have no use whatsoever.
false. While it is true that naive forecasts don't tend to be highly accurate, we can use their relative inaccuracy as a benchmark to measure other forecasts against. If they don't do as well as the naive, then they may not be appropriate for the data in question.
Qualitative methods utilizes quantitative data.
false. While qualitative methods can use the results of a survey, these are not considered quantitative data because a survey elicit opinions rather than facts.
Naive forecasts are simple to compute: we write the formula as Ft= A t
false. While they are simple, the formula is Ft= A t-1 Remember, without clairvoyance we can only forecast given prior data.
All forecasts should look out at least 3 months.
false: It depends on the forecast's purpose. A forecast done to help us figure out how much to order to restock our shelves on a weekly basis doesn't need to look so far out (unless we are in the unlikely situation where we can't reorder again for several months!)
Francois decides to compare his Frose' models after a few more days of sales. It turns out the correlation for linear trend analysis is .65 and the one for associative forecast is .72. Use what appears to be the more accurate model to predict sales on day 12, when the temperature is expected to only get to 70 because the heat wave is dissapating.
he associative forecast appears to be more accurate, as the correlation is higher (.72> .65). We use: Yt = -40+ 1.5*F (where F is degrees Farenheit) for F = 70, the model predicts -40+1.5*70 = 65 sales of Frose' Francois is saddened at the drop in expected sales as well as the arrival of cooler temperatures. But he realizes life could be worse: he could be stuck in Akron selling space heaters during an Midwestern winter.
Forecasting are more accurate in the near future than the distant future.
true
Qualitative methods are subjective methods and incorporate opinions into the forecast.
true.
Qualitative methods can be useful in predicting sales of a new product.
true.
The simple naive forecast requires only one data point.
true.
It's important to plot the data and check for patterns and outliers before settling on a forecasting technique.
true. As we will see later, if there's a trend, we don't immediately want to go to a smoothing filter
