Predictive Statistics Test Selection Help

hockeydata16 — Wed, 07 Apr 2021 20:18:06 GMT

Hello,

I am needing help determining which are the appropriate predictive statistical tests to run to prepare my dataset for predicting future sales for each location. The dataset consists of 24 observations, a month column (independent variable), and 4 different City columns (dependent variables). The values recorded for each city during the observation is the total sales (in dollars).

The dataset essentially looks like:

Month Dallas Boston NYC Chicago

1 34245 58745 62145 12345

.......

24 87451 58472 21451 64125

I think a correlation test will be needed to see if there is a relationship between any of the cities and then finally a linear regression model to predict future values of sales at each location. thank you in advance for your assistance!

Re: Predictive Statistics Test Selection Help

PaigeMiller — Wed, 07 Apr 2021 21:13:55 GMT

This seems to me like any prediction ought to take into account the auto-correlation between the different time points, and so linear regression would not work well if there is such auto-correlation. You probably ought to consider time-series modeling, if you have SAS/ETS then there are a number of procedures that can perform such modeling, including PROC ARIMA.

topic Re: Predictive Statistics Test Selection Help in Statistical Procedures

Predictive Statistics Test Selection Help

Re: Predictive Statistics Test Selection Help