Introduction to Data Analytics 2 Part 1: Experiment Design
Title: Consumer Pseudo-Showrooming and Omni-Channel Product Placement Strategies
Abstract: Recent advances in information technologies (IT) have powered the merger of online and offline retail channels into one single platform. Modern consumers frequently switch between online and offline channels when they navigate through various stages of the decision journey, motivating multichannel sellers to develop omni-channel strategies that optimize their overall profit. This study examines consumers’ cross-channel search behavior of "pseudo-showrooming," or the consumer behavior of inspecting one product at a seller’s physical store before buying a related but different product at the same seller’s online store, and investigates how such consumer behavior allows a multichannel seller to achieve better coordination between its online and offline arms through optimal product placement strategies.
Participants in the study were grouped into the following categories:
A. Where_bought: Where they ended up purchasing an item: bought at the store, bought online.
B. Who_bought: If they bought from the same or a different retailer.
Each participant was then measured on:
A. Money: how much money they spent in dollars on the product.
B. Time: how much time (in minutes) they spent looking at the product online.
1. What would be one possible null hypothesis based on this study? The time that a consumer spent looking at the product online does not have a relationship with their decision to buy from the same or a different retailer.
2. What would be one possible alternative hypothesis based on this study? The time that a consumer spent looking at the product online has a relationship with their decision to buy from the same or a different retailer.
3. Who are they sampling in this study? Consumers
4. Who is the intended population in this study? Modern consumers who making purchases either online or offline.
5. Give an example of type 1 error based on this study (do not just define, explain in context how it might have happened here). A type 1 error occurs when detecting an effect that is not present. In this case, the type 1 error could be: the amount of money a consumer spent on a product does not have a relationship with if they bought it from the same retailer or not, but this null hypothesis is rejected based on bad data analysis.
6. Give an example of type 2 error based on this study (do not just define, explain in context how it might have happened here). A type 2 error occurs when failing to detect an effect that is present. In this case, the type 2 error could be: the amount of time a consumer spent on looking at a product online is related to the decision of where to buy this product, but the data are such that the null hypothesis cannot be rejected.
Part 2: Use the 04_data. csv to complete this portion.
1. For each IV list the levels (next to a, b):
A. Where bought: store vs. online
B. Who bought:same vs. different
2. What are the conditions in this experiment? Bought in store from same retailer; Bought in store from different retailer; Bought online from same retailer; Bought online from different retailer.
3. For each condition list the means, standard deviations, and standard error for the conditions for time and money spent. Please note that means you should have several sets of M, SD, and SE. Be sure you name the sets of means, sd, and se different things so you can use them later.
4. Which condition appears to have the best model fit using the mean as the model (i. e. smallest error) for time? Bought in store from different retailer.
5. What are the df for each condition?
6. What is the confidence interval (95%) for the means?
##money-M1
##time-M2
7. Use the MOTE library to calculate the effect size for the difference between money spent for the following comparisons (that means you’ll have to do this twice):
##Store versus online when bought at the same retailer
##Store versus online when bought at a different retailer
8. What can you determine about the effect size in the experiment - is it small, medium or large? I would say in terms of number of observations of 40, it’s quite small. For the money they spend, there isn’t much difference, so it’s small. For the time we can observe some difference so it’s medium
9. How many people did we need in the study for each comparison?
#Two-sample t test power calculation n = 512.2686d = 0.1752176sig. level = 0.05power = 0.8alternative = two. sidedNOTE: n is #number in *each* grou