The foregoing scheme of standardization has been submitted to a number of hygienists and economists, and, while there is a variety of opinion regarding details, there is a striking unanimity as to the need and value of standardization itself. If the matter were to be taken up in a comprehensive way, there is no reason why standardization should not be established in hygiene and economics as it is in agriculture and engineering.

DEVICES FOR AVOIDING ERROR.

BY GEORGE K. HOLMES.

A deal of drudgery should accompany the tabulation of statistical material, the preparation of printer's copy derived therefrom, and the verification of proofs. During many years of contact with professors and students of colleges and universities, with officials of State bureaus of labor, and with persons, official and unofficial, engaged in statistical work, the writer has often observed a want of safeguards against error. Instead of assuming, as should be assumed, that every conclusion is wrong until fortified by verification, it is not infrequently the habit of workers to regard results as true if obtained by themselves. What statistician of experience has not heard a clerk say, "I am sure that this is right, because I did it myself," or "I never make an error"? Such clerks are the most dangerous ones in a statistical office, and are not to be trusted at any time or for any purpose, unless under precautions beyond their control or even knowledge.

A common method of verifying a copy of a statistical table and of text relating thereto, as well as of verifying a proof, is to depend solely upon reading back to copy. In consequence of this, publications, whether statistical or otherwise and whether governmental or of other origin, contain many errors. Indeed, some publications are really "honey-combed" with errors. The report of the census of agriculture of a certain. State contains a series of tables in which there is a presentation of number of acres and numbers of units of production of principal crops by counties,-simple statements, covering no great number of pages. The accuracy of the publication being called in question, an attempt to prove the additions resulted in the discovery of many hundreds of errors on the face of the tabu

lar statements. No one familiar with precautions against errors in statistical work could have any confidence in the accuracy of the work that produced those tables.

It is a common observation in the use of statistical matter published by State bureaus of labor, boards of agriculture, other State offices, and, indeed, by some national offices, that there has been a failure to follow a safe course of procedure in office work and in verifying printed matter, and the published productions of college and university professors and students are far from being exempt from this characterization.

The observation of the writer has been that professors and graduate students of universities and colleges possess a mind trained to establish facts, but one almost without precautions to preserve facts in their transfer from the primary source to the stereotyped printed page. It is no attempt at exaggeration to say that a girl graduated from a high school not longer than one or two years previously can better be trusted to safeguard the transfer than they can.

A common sole reliance for the verification of a copy or of a proof is a reading back to the original. This proceeding is fraught with errors. While it is wise to read back to copy or to original to discover the bulk of the errors, a sole dependence upon this proceeding is sure to leave many errors undiscovered. Such being the general situation, an excuse is offered for the presentation of a formulated description of precautions against error in the statement that follows. These are based upon the many years' experience of the writer and of other persons engaged in statistical work. They may be depended upon to eliminate error almost completely. The qualification "almost completely" is made because there are two or three sources of error which no prevision can avoid, namely: (1) the concerted dishonesty of workers; (2) compensating errors, as when a plus error is counterbalanced by a minus error,-a very rare occurrence, yet one which may and does occur; and (3) the repetition of an error in the same place by different persons working at different times and independently of each other. The last error is exceedingly rare, and yet it does occur. An instance

is called to mind in which a series of percentages were computed, all proving to 100 after justification. Three persons performed this work, the results of each being unknown to the others, and the results were the same in every particular; yet a fourth person repeating the operation discovered a common error which the preceding three had committed. An experience of this sort might not be repeated probably in a lifetime.

For the use of practical workers in statistics, whether in educational institutions or in government offices, who are relying solely upon their own work and upon reading copies and proofs back to originals, the following is offered:

Table-making:

RULES TO GOVERN WORK.

Table-making generally to be in duplicate when feasible; when not, the figures, additions, and computations to be verified by a second worker. When tables are made in duplicate or are to be copied, having columns without totals, the columns are to be added to get totals for "check" purposes.

In comparing duplicate tables, or in verifying a table, or in verifying a copy of a table or of text, do not erase a figure supposed to be wrong: use red ink to correct it above.

Use red ink for total lines.

Persons who do duplicate work, or first work and its verification, are not to confer with each other concerning differences nor to consult with each other before finishing work.

The initials of each person doing work should appear at an upper corner of the first sheet, together with a statement as to what work was done: thus, "Compiled," "Verified," "Compared," etc.

Making and verifying copies of tables:

Do not copy total lines; get totals by adding copy and compare with original.

If original has no total lines, add to get them for "check" purposes; also add the copy, and compare.

Read copy back to original before adding.

Text "copy":

To be verified from accompanying tables or from original sources.

A copy of this to be verified by at least one reading back and by procedure of preceding paragraph.

First galley proof of tables and text:

Read critically, and make thereon all superficial corrections.

Copy on retained proof or elsewhere for future use the "check" totals of tables of printer's copy.

First page proof of tables:

See that all corrections indicated on galley proof have been made.
Add columns to prove to total lines and to "check" totals.
Cross-add to total columns.

Make all computations that can be made from the proof.

First page proof of text:

See that all corrections indicated on galley proof have been made.
Make all computations that can be made from the proof.

Verify from accompanying tables and from printer's "copy."

Second page proof of tables and text:

Read last column of tables back to copy, and add. (Types will fall down and not get back to their proper places.)

See that all corrections indicated on first page proof have been made.

Departures from the foregoing are permissible for special and good reasons, depending on the importance of the work, the ability and honesty of clerks, and the practical difficulties that are encountered; but not at the discretion of any one but the person in charge.