Adverse weather conditions are known to cause or contribute to power outages. Power utility companies are interested in knowing the amount of power outages based on future weather conditions for the sake of saving time and money. This study uses power outage data from Public Service of New Hampshire (PSNH) spanning the years 2006-2010 to create a power outage event definition. Using the power outage event definition, statistical analyses were performed using histograms, boxplots, and regressions. The weather variables used in this study were temperature, dewpoint temperature, wind speed, wind gust, relative humidity, altimeter, visibility, and ceiling height. Cloud cover and present weather were also included. Histograms showed that it was difficult to distinguish between events and nonevents using PSNH variables(number of Trouble Report and Unsatisfactory Performance of Equipment Reports(TRUPERs)per event, length of outage event, customers, customer min, and duration)and weather variables. The distributions of the sum of each PSNH variable were similar for events and nonevents. The mean and median values of each PSNH variable were higher in events compared to nonevents. The distributions for the mean, median, maximum, and minimum values of each weather variable were similar for events and nonevents. The distributions for the number of occurrences for mean wind speeds and wind gusts along with minimum visibility, altimeter, and ceiling height suggested events may be distinguished from nonevents using these variables. Events were split into warm and cold season events and then further into large and small events. Boxplots were constructed for all, large, and small events in the warm and cold season for each meteorological variable. Warm season large events (WSLEs)had warmer maximum temperatures and dewpoint temperatures, higher maximum wind speeds and wind gusts, and lower minimum altimeter and visibilities compared to warm season small events (WSSEs). Cold season large events (CSLEs)had warmer maximum dewpoint temperatures, higher maximum wind speeds and wind gusts, and lower minimum altimeter and visibilities compared to cold season small events (CSSEs). Three sets of regression equations were constructed. The first set of equations consisted of using variables that had |correlation| ≥0.3 to predict each of the PSNH variables. The second set of equations were a trimmed version of the first set of regression equations. The third set of equations were created using a stepwise regression technique. Contingency tables were constructed to assess the performance of each regression equation. The probability of detection (POD), false alarm rate (FAR), probability of false detection (POFD), and combined skill index (CSI) were then computed for each regression equation to assess how well each regression equation was at predicting whether a large event would occur. The stepwise regressions outperformed the other regressions in predicting whether or not a large event would occur. Predicting the number of TRUPERs in an event showed the most promise using the stepwise regressions. Overall, predicting an event from a nonevent or a large event from a small event is difficult when using the weather variables in this study.