August 18, 2009

Question:

Scott Dermont, IA

I have a library whose server crashed and lost all of their data between 2005 and April 2009. I’m wondering what I should tell them as far as filling out their FY09 survey? Should I just say to skip it for this year and wait until they have a full year’s data? Or should I have them extrapolate off of the data that they managed to collect for 3 months? I didn’t see anything like this on the wiki so I thought I’d email the group for advice.

Thanks in advance for any help that you can provide.

IMLS Response:

After talking with the Library Statistics Working Group, the consensus was that the library should extrapolate using the three months of data available and use the past couple of submissions as a check that the numbers they’re getting are reasonable.

Also, since this is a relatively small library, when it is practical for them to manually count elements such as the number of staff, or obtain budgetary/fiscal figures from other sources, we ask that you do so (this last part is just practical advice from IMLS and Census).

Comments:

Susan Mark, WY

Were it one of mine, I'd probably suggest that they make the best estimate they can using the three months of data they have left, and the prior year's data. I don't know if that's the correct thing to do, but it's what I would probably do.

Howard Boksenbaum, RI

I would suggest a straight line extrapolation as coming as close as possible.  Is there a way to indicate that these data were estimated on the final submission?

Michael Golrick, LA

As you may remember, Louisiana (along with Mississippi, Alabama, Florida, and Texas) have had the experience of dealing with missing data with whole libraries blown away (literally). One parish in here (Cameron) literally had nothing left after Katrina.

Scott did tell me in a private note that the library had been doing back-ups, but they were not sufficient. I was going to make a snarky comment, but won’t. I will say that as a public library director, and even with the back up procedures, there were some records (board minutes including financial reports and statistics) which were also put into print and, in one library, bound as a permanent record.

Now, I was not here in 2005, and with the number of issues resulting from two hurricanes, the institutional memory of how the statistics were handled is fuzzy. I did go and pull the state published report from that time. There is data for all libraries, so some of it must have been reconstructed and/or estimated. There is data in our reports for even the most heavily hit libraries, which makes me think that there was estimating done.

I expect that the official advice will be to estimate in the most reasonable manner possible.

Stacey Malek, TX

In Texas, we ask libraries affected by hurricane damage to simply do the best they can with reporting their statistics, including estimating.  They put annotations into bibiostat explaining why numbers were off between years.

Edie Huffman, IN

We've done the same in Indiana.  One of our libraries was destroyed by fire (arson, and the new library was only 2 years old), so all of their computer and most print records were gone. 

Timothy Owens, NC

It seems like they should be able to treat data they have (or a fraction thereof) as a sample and be able to produce some meaningful numbers. Carlos discussed sampling at the conference last year and may have some ideas on a practical approach.

Post IMLS Response:

Scott Dermont, IA

I wanted to thank everyone who helped me with the server crash question. The consensus was to extrapolate, which is very logical to me. Everett from IMLS also followed up with me about this and also recommended extrapolation.  Thanks again.