Statistical schedules are filled out in vari ous ways. Census blanks are taken from house to house by paid enumerators who write in the information obtained by direct questioning. They are instructed to make all necessary ex, planation and to require answers to the ques tions asked, as refusal to answer is made a misdemeanor. Thus little difficulty is experi enced in securing complete data. Statistics of births and deaths are usually obtained from schedules filled out by attending physicians. Statistics of inmates of institutions are best prepared by having uniform schedules filled out and sent to a central statistical office. Elab orate statistical reports on prepared schedules are obtained by the Interstate Commerce Com mission and by public service commissions from companies operating public utilities. In order to make the reports comparable, uniform sys. tems of accounting have been devised for com panies of the various classes. The data now being collected from such companies are note worthy examples of the efficient use of statis tical methods. Labor bureaus secure data rela tive to wages and employment by sending ques tionnaires to employers and to labor unions and by means of special agents. Social and philanthropic organizations make special sta tistical studies or surveys by the use of ques tionnaires sent out by mail or by means of investigations made by field workers; the lat ter although of limited scope is usually more successful. When the schedules have been filled out and returned to the statistical office they must be examined for errors and incon sistencies and corrected before tabulation is undertaken. In spite of every precaution some mistake§ will occur on the schedules and some desired data will be unascertained.
Two methods are used in presenting sta tistical data to the general reader, namely, tabulation and charting; the former is the pri mary method and is almost universal; the lat ter is supplemental and of limited use. The preparation of effective statistical tables re quires food judgment and a thorough under standiifg of the significance of the data and of the questions that are to be answered by the tables. Although the numerical table is an end in statistical work the plan of tabulation must be decided upon before the schedule is prepared. The latter must be so formulated that it will furnish the information needed for the table. Tabulations should be as simple as the data will permit and should show only significant aggregates. The necessity of making the data comparable with corresponding data in previous studies must be borne in mind. Tabulations may be made either by sorting by hand and counting or by machines specially devised for such purpose. The number of cards and the kinds of tabulations to be made determine the method to be used. In census work, where elaborate studies of millions of individuals are made, machines are indispens able. Machine tabulation necessitates the use
of codes whereby the data on the schedule card may be transferred to the card to be used in the machine by punching holes at prescribed places. The punching machines and cards are made to meet the requirements of the sorting and tabulating machines. The electric tabulat ing machines will make cross tabulations at the rate of 250 to 400 cards per minute.
Statistical charts or graphs are used to il lustrate statistical tables. They make clear and emphasize the principal numerical results shown in the tables. To be effective, statistical charts must meet certain definite requirements: (a) They must be self-explanatory. Charts are useless if they can be understood only after reading lengthy explanations in the text or footnotes. (b) They must be simple. Com plicated charts are more difficult to understand than statistical tables, and therefore fail in their purpose. (c) They must have unity, i.e., not more than one principal theme should be por trayed on a single chart. A large variety of graphic forms are used in making charts, the most common of which are points, bars, curves, areas and maps. The kind of statistical results to be shown determines the form to be used in any case. To standardize the use of diagrams, a joint committee on standards for graphic presentation representing 17 national societies was appointed in 1915. A preliminary report of this committee sets forth the following prin ciples as applicable to elementary graphic pre sentation: The general arrangement of a diagram should proceed from left to right.
Where possible, represent quantities by linear magni tudes, as areas or volumes are more likely to be misin terpreted.
For a curve the vertical wale, whenever practicable, should be so selected that the zero line will appear on the diagram.
If the zero line of the vertical scale will not normally appear on the curve diagram, the zero line should be shown by the use of a horizontal break in the diagram.
The zero lines of the scales for a curve should be sharply distinguished from the other co-ordiratte tines: For curves having a scale representing percentages, it is usually desirable to emphasise in some distinctive way the line or other line used as a basis of compafison.
the scale of a diagram refers to dates, and the period represented is not a complete unit, it is better not to emphasize the first and last ordinates, since such a diagram does not represent the beginning or end of time.
When curves are drawn on logarithmic co-ordinates, the lines of the diagram should each be at some power of 10 on the logarithmic scales.
It is advisable not to show any more co-ordinate lines than necessary to guide the eye in reading the diagram.
The curve lines of a diagram should be sharply dis tinguished from the ruling.